Upstream updates detected — 2026-05-27

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: @upstash/context7-pi@0.1.0 (was ctx7@0.4.2)
• Release notes:

### Minor Changes

-   f91b40c: Initial release. Adds an official Context7 extension for the [pi coding agent](https://pi.dev) — registers `resolve-library-id` and `query-docs` tools, ships the `context7-docs` skill, and exposes a `/c7-docs` slash command. Wire format, error messages, and tool descriptions are copied verbatim from `@upstash/context7-mcp` so pi and MCP clients give the LLM identical instructions and output. Self-contained — no Context7 runtime dependencies. Works out of the box at IP-based rate limits; set `CONTEXT7_API_KEY` for the higher tier. Install with `pi install npm:@upstash/context7-pi`.

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 2 docs · 4 feat · 1 fix · 2 other
• Files touched: code=37 docs=12 ci=0 test=3 other=2
• 12 new commit(s) since last check
• Latest: 7e956e5 — chore(release): version packages (#2675)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• New release: v1.5.0rc3 (was v1.4.16)
• Release notes:

## Major Improvements

*  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  The file processing pipeline supports selecting the content parsing engine and text chunking strategy either based on file extensions or on a per-file basis. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Add optional **JSON-formatted LLM** output to enhance stability in the entity and relation extraction pipeline.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

<img width="945" height="679" alt="image" src="https://github.com/user-attachments/assets/b5307383-ac00-4f98-9821-061b6e0cee00" />

## What's Broken

* **The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.

• Impact: 🟡 review (code touched)
• Commit mix: 3 build · 1 docs · 2 feat · 11 fix · 230 other · 2 refactor · 1 revert
• Files touched: code=0 docs=25 ci=11 test=132 other=133
• 760 new commit(s) since last check
• Latest: b313eb4 — 📝 docs(ProgramingWithCore): expand addon_params documentation
• Recent commits:
• 6b5a086 [other] Merge pull request #3081 from danielaskdd/claude/mineru-page-positions
• 4878908 [other] 📝 docs(FileProcessingConfiguration-zh): expand mineru and docling setup instructions
• fde8539 [other] Tokenizer.encode: gracefully handle disallowed special tokens
• 863d03e [other] write_nx_graph: atomic write via per-writer .tmp + os.replace
• 88f6dac [other] 📝 docs(sidecar): update chinese spec and add per-position origin and self_ref fields

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 10 other
• Files touched: code=115 docs=70 ci=0 test=65 other=93
• 33 new commit(s) since last check
• Latest: 1801156 — Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 31 docs · 8 feat · 18 fix · 8 other · 6 test
• Files touched: code=0 docs=71 ci=4 test=25 other=51
• 82 new commit(s) since last check
• Latest: 928076c — chore: export marketing campaign command
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

n8n-mcp

• New release: v2.56.0 (was v2.53.2)
• Release notes:

# Release v2.56.0

Generating release notes from v2.55.0 to HEAD
### 🔧 Maintenance

- Bump version to 2.56.0 and add changelog entry for additionalTools hook (a0d1e08)

### 📝 Other Changes

- refactor(mcp): final cleanups from review on additionalTools hook (afca30d)
- fix(mcp): security and consistency hardening for additionalTools (7699b92)
- Address review feedback: type fix, telemetry docs, new tests, dist cleanup (#801) (789ff80)
- Align additionalTools execution with core `tools/call` semantics (#800) (76228db)
- Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en) (906d6bf)
- Initial plan (afdad8b)

---

**Release Statistics:**
- 7 commits

• Impact: 🟢 skim (docs/ci/test only)
• Files touched: code=0 docs=0 ci=0 test=0 other=0
• ? new commit(s) since last check
• Latest: 0f3d3f5 — Merge pull request #799 from czlonkowski/copilot/add-additional-tools-hook

claude-mem

• New release: v13.3.0 (was v13.2.0)
• Release notes:

## What's New

### New skills

- **design-is** (#2483) — audits a design against Dieter Rams' ten "Good design is..." principles. Produces per-principle 0–3 scores with file:line evidence and a NEW / REFINE / REDESIGN verdict, then hands off a ready-to-run `/make-plan` prompt.
- **weekly-digests** (#2399) — produces a chapter-per-ISO-week serial digest of a project's full claude-mem timeline. Sequential subagent pipeline keeps the narrative coherent across 30+ chapters.
- **oh-my-issues** (#2409) — root-cause issue clustering. Codifies the consolidation method that turned ~100 open issues into 6 plan-masters during the v13.0.1 cycle. Three modes: cluster pass, triage, bundle.

### Fixes

- **fix(mcp): drop duplicate root `.mcp.json`** (#2411) — Claude Code's `/doctor` was warning "MCP server mcp-search skipped — same command/URL as already-configured mcp-search" for every plugin user. The root copy was vestigial; the plugin's namespaced registration now wins.
- **fix: stop Codex transcript replay after hooks migration** (#2365) — disables the default `~/.codex/sessions/**/*.jsonl` watch (native Codex hooks are now authoritative). Repairs `~/.codex/config.toml` to set `[features] hooks = true` and `[plugins."claude-mem@claude-mem-local"] enabled = true` directly. Fixes transcript replay where files discovered after startup ignored `startAtEnd` and re-injected history.

Opt back into legacy Codex transcript ingestion with `CLAUDE_MEM_CODEX_TRANSCRIPT_INGESTION=true` if you depend on the JSONL watcher.

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 1 docs · 3 feat · 2 fix
• Files touched: code=6 docs=5 ci=0 test=5 other=16
• 7 new commit(s) since last check
• Latest: c3d2af7 — docs: regenerate CHANGELOG for v13.3.0
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• Impact: 🟢 skim (docs/ci/test only)
• Files touched: code=0 docs=0 ci=0 test=0 other=0
• ? new commit(s) since last check
• Latest: d0a7629 — Merge branch 'chore/dependabot-group-all-bumps'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: @upstash/context7-pi@0.1.0 (was ctx7@0.4.2)
• Release notes:

### Minor Changes

-   f91b40c: Initial release. Adds an official Context7 extension for the [pi coding agent](https://pi.dev) — registers `resolve-library-id` and `query-docs` tools, ships the `context7-docs` skill, and exposes a `/c7-docs` slash command. Wire format, error messages, and tool descriptions are copied verbatim from `@upstash/context7-mcp` so pi and MCP clients give the LLM identical instructions and output. Self-contained — no Context7 runtime dependencies. Works out of the box at IP-based rate limits; set `CONTEXT7_API_KEY` for the higher tier. Install with `pi install npm:@upstash/context7-pi`.

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 2 docs · 4 feat · 1 fix · 2 other
• Files touched: code=37 docs=12 ci=0 test=3 other=2
• 12 new commit(s) since last check
• Latest: 7e956e5 — chore(release): version packages (#2675)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• New release: v1.5.0rc3 (was v1.4.16)
• Release notes:

## Major Improvements

*  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  The file processing pipeline supports selecting the content parsing engine and text chunking strategy either based on file extensions or on a per-file basis. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Add optional **JSON-formatted LLM** output to enhance stability in the entity and relation extraction pipeline.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

<img width="945" height="679" alt="image" src="https://github.com/user-attachments/assets/b5307383-ac00-4f98-9821-061b6e0cee00" />

## What's Broken

* **The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.

• Impact: 🟡 review (code touched)
• Commit mix: 3 build · 2 docs · 3 feat · 16 fix · 214 other · 6 perf · 4 refactor · 1 revert · 1 test
• Files touched: code=0 docs=25 ci=11 test=131 other=134
• 806 new commit(s) since last check
• Latest: a616d8e — 📝 docs(milvus): clarify drop method caveats
• Recent commits:
• 3bf2297 [other] Merge pull request #3082 from RooseveltAdvisors/tokenizer-disallowed-special-fallback
• 1377158 [other] ♻️ refactor(mineru): replace sync file reads with streaming upload for httpx 0.28+ compatibility
• 29dbaf9 [other] Merge pull request #3089 from danielaskdd/refact/mineru-enhance
• 224f85b [other] Merge branch 'main' into dev
• 4224800 [other] ♻️ refactor(constants): centralize parsed artifact directory suffixes

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 10 other
• Files touched: code=115 docs=70 ci=0 test=65 other=93
• 33 new commit(s) since last check
• Latest: 1801156 — Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 32 docs · 8 feat · 18 fix · 9 other · 6 test
• Files touched: code=0 docs=88 ci=5 test=25 other=55
• 84 new commit(s) since last check
• Latest: 04c68e4 — Add React language track with agents, skills, rules, and commands (#2024)
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

n8n-mcp

• New release: v2.56.0 (was v2.53.2)
• Release notes:

# Release v2.56.0

Generating release notes from v2.55.0 to HEAD
### 🔧 Maintenance

- Bump version to 2.56.0 and add changelog entry for additionalTools hook (a0d1e08)

### 📝 Other Changes

- refactor(mcp): final cleanups from review on additionalTools hook (afca30d)
- fix(mcp): security and consistency hardening for additionalTools (7699b92)
- Address review feedback: type fix, telemetry docs, new tests, dist cleanup (#801) (789ff80)
- Align additionalTools execution with core `tools/call` semantics (#800) (76228db)
- Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en) (906d6bf)
- Initial plan (afdad8b)

---

**Release Statistics:**
- 7 commits

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 1 ci · 1 feat · 1 fix · 5 other · 1 refactor
• Files touched: code=9 docs=39 ci=2 test=2 other=39
• 11 new commit(s) since last check
• Latest: 0f3d3f5 — Merge pull request #799 from czlonkowski/copilot/add-additional-tools-hook
• Recent commits:
• f1edbc9 [feat] feat(mcp): serve n8n-skills markdown via MCP Resources (#793)
• dc272cf [ci] ci(release): switch npm publishing to Trusted Publishers (OIDC) (#796)
• 41e71f8 [chore] chore: update n8n to 2.21.7 and bump version to 2.55.0 (#797)
• afdad8b [other] Initial plan
• 906d6bf [other] Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en)

claude-mem

• New release: v13.3.0 (was v13.2.0)
• Release notes:

## What's New

### New skills

- **design-is** (#2483) — audits a design against Dieter Rams' ten "Good design is..." principles. Produces per-principle 0–3 scores with file:line evidence and a NEW / REFINE / REDESIGN verdict, then hands off a ready-to-run `/make-plan` prompt.
- **weekly-digests** (#2399) — produces a chapter-per-ISO-week serial digest of a project's full claude-mem timeline. Sequential subagent pipeline keeps the narrative coherent across 30+ chapters.
- **oh-my-issues** (#2409) — root-cause issue clustering. Codifies the consolidation method that turned ~100 open issues into 6 plan-masters during the v13.0.1 cycle. Three modes: cluster pass, triage, bundle.

### Fixes

- **fix(mcp): drop duplicate root `.mcp.json`** (#2411) — Claude Code's `/doctor` was warning "MCP server mcp-search skipped — same command/URL as already-configured mcp-search" for every plugin user. The root copy was vestigial; the plugin's namespaced registration now wins.
- **fix: stop Codex transcript replay after hooks migration** (#2365) — disables the default `~/.codex/sessions/**/*.jsonl` watch (native Codex hooks are now authoritative). Repairs `~/.codex/config.toml` to set `[features] hooks = true` and `[plugins."claude-mem@claude-mem-local"] enabled = true` directly. Fixes transcript replay where files discovered after startup ignored `startAtEnd` and re-injected history.

Opt back into legacy Codex transcript ingestion with `CLAUDE_MEM_CODEX_TRANSCRIPT_INGESTION=true` if you depend on the JSONL watcher.

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 1 docs · 3 feat · 2 fix
• Files touched: code=6 docs=5 ci=0 test=5 other=16
• 7 new commit(s) since last check
• Latest: c3d2af7 — docs: regenerate CHANGELOG for v13.3.0
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 2 fix · 16 other · 2 refactor · 3 test
• Files touched: code=0 docs=7 ci=1 test=3 other=11
• 31 new commit(s) since last check
• Latest: f6c60e4 — Merge branch 'docs/VM-1400-surface-permission-allow-list-in-voicemode-skill'
• Recent commits:
• 0855fbd [other] install.sh: install + configure mlx-audio on Apple Silicon (VM-1330)
• 7646167 [other] install.sh: offer mlx-audio on Apple Silicon even if whisper+Kokoro present (VM-1330)
• bb1b666 [other] install.sh: status-first voice-services flow, skip prompt when satisfied, reworded prompt, de-emoji banner (VM-1330)
• 073fada [other] .mcp.json: invoke launcher as bare command, not uv run (VM-1340)
• bb0f68d [other] Merge branch 'research/VM-1340-resolve-plugin-mcpjson-launcher-invocation-bare'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: @upstash/context7-pi@0.1.0 (was ctx7@0.4.2)
• Release notes:

### Minor Changes

-   f91b40c: Initial release. Adds an official Context7 extension for the [pi coding agent](https://pi.dev) — registers `resolve-library-id` and `query-docs` tools, ships the `context7-docs` skill, and exposes a `/c7-docs` slash command. Wire format, error messages, and tool descriptions are copied verbatim from `@upstash/context7-mcp` so pi and MCP clients give the LLM identical instructions and output. Self-contained — no Context7 runtime dependencies. Works out of the box at IP-based rate limits; set `CONTEXT7_API_KEY` for the higher tier. Install with `pi install npm:@upstash/context7-pi`.

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 2 docs · 4 feat · 1 fix · 2 other
• Files touched: code=37 docs=12 ci=0 test=3 other=2
• 12 new commit(s) since last check
• Latest: 7e956e5 — chore(release): version packages (#2675)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• New release: v1.5.0rc3 (was v1.4.16)
• Release notes:

## Major Improvements

*  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  The file processing pipeline supports selecting the content parsing engine and text chunking strategy either based on file extensions or on a per-file basis. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Add optional **JSON-formatted LLM** output to enhance stability in the entity and relation extraction pipeline.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

<img width="945" height="679" alt="image" src="https://github.com/user-attachments/assets/b5307383-ac00-4f98-9821-061b6e0cee00" />

## What's Broken

* **The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.

• Impact: 🟡 review (code touched)
• Commit mix: 7 build · 2 docs · 4 feat · 17 fix · 208 other · 7 perf · 3 refactor · 1 revert · 1 test
• Files touched: code=0 docs=25 ci=11 test=129 other=136
• 855 new commit(s) since last check
• Latest: 57f9116 — Update env.exmaple
• Recent commits:
• 46576b2 [other] 🔧 chore(docker): add prompt directory configuration across docker and setup scripts
• 69aed36 [other] 🔧 refactor(logging): unify and simplify pipeline log messages
• ac025b1 [other] ♻️ refactor(doc-status-storage): convert concrete methods to abstract methods
• f3dddb1 [other] 📝 docs(docs): update log message in file processing pipeline documentation
• c050cc4 [other] 🔧 chore(gitignore): update ignore rules for prompts directory

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 10 other
• Files touched: code=115 docs=70 ci=0 test=65 other=93
• 33 new commit(s) since last check
• Latest: 1801156 — Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 32 docs · 8 feat · 19 fix · 9 other · 6 test
• Files touched: code=0 docs=88 ci=5 test=28 other=61
• 85 new commit(s) since last check
• Latest: 64cd1ba — fix: surface warn-only PreToolUse hooks (#2084)
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

n8n-mcp

• New release: v2.56.0 (was v2.53.2)
• Release notes:

# Release v2.56.0

Generating release notes from v2.55.0 to HEAD
### 🔧 Maintenance

- Bump version to 2.56.0 and add changelog entry for additionalTools hook (a0d1e08)

### 📝 Other Changes

- refactor(mcp): final cleanups from review on additionalTools hook (afca30d)
- fix(mcp): security and consistency hardening for additionalTools (7699b92)
- Address review feedback: type fix, telemetry docs, new tests, dist cleanup (#801) (789ff80)
- Align additionalTools execution with core `tools/call` semantics (#800) (76228db)
- Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en) (906d6bf)
- Initial plan (afdad8b)

---

**Release Statistics:**
- 7 commits

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 1 ci · 1 feat · 1 fix · 5 other · 1 refactor
• Files touched: code=9 docs=39 ci=2 test=2 other=39
• 11 new commit(s) since last check
• Latest: 0f3d3f5 — Merge pull request #799 from czlonkowski/copilot/add-additional-tools-hook
• Recent commits:
• f1edbc9 [feat] feat(mcp): serve n8n-skills markdown via MCP Resources (#793)
• dc272cf [ci] ci(release): switch npm publishing to Trusted Publishers (OIDC) (#796)
• 41e71f8 [chore] chore: update n8n to 2.21.7 and bump version to 2.55.0 (#797)
• afdad8b [other] Initial plan
• 906d6bf [other] Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en)

claude-mem

• New release: v13.3.0 (was v13.2.0)
• Release notes:

## What's New

### New skills

- **design-is** (#2483) — audits a design against Dieter Rams' ten "Good design is..." principles. Produces per-principle 0–3 scores with file:line evidence and a NEW / REFINE / REDESIGN verdict, then hands off a ready-to-run `/make-plan` prompt.
- **weekly-digests** (#2399) — produces a chapter-per-ISO-week serial digest of a project's full claude-mem timeline. Sequential subagent pipeline keeps the narrative coherent across 30+ chapters.
- **oh-my-issues** (#2409) — root-cause issue clustering. Codifies the consolidation method that turned ~100 open issues into 6 plan-masters during the v13.0.1 cycle. Three modes: cluster pass, triage, bundle.

### Fixes

- **fix(mcp): drop duplicate root `.mcp.json`** (#2411) — Claude Code's `/doctor` was warning "MCP server mcp-search skipped — same command/URL as already-configured mcp-search" for every plugin user. The root copy was vestigial; the plugin's namespaced registration now wins.
- **fix: stop Codex transcript replay after hooks migration** (#2365) — disables the default `~/.codex/sessions/**/*.jsonl` watch (native Codex hooks are now authoritative). Repairs `~/.codex/config.toml` to set `[features] hooks = true` and `[plugins."claude-mem@claude-mem-local"] enabled = true` directly. Fixes transcript replay where files discovered after startup ignored `startAtEnd` and re-injected history.

Opt back into legacy Codex transcript ingestion with `CLAUDE_MEM_CODEX_TRANSCRIPT_INGESTION=true` if you depend on the JSONL watcher.

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 1 docs · 3 feat · 2 fix
• Files touched: code=6 docs=5 ci=0 test=5 other=16
• 7 new commit(s) since last check
• Latest: c3d2af7 — docs: regenerate CHANGELOG for v13.3.0
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 2 fix · 16 other · 2 refactor · 3 test
• Files touched: code=0 docs=7 ci=1 test=3 other=11
• 31 new commit(s) since last check
• Latest: f6c60e4 — Merge branch 'docs/VM-1400-surface-permission-allow-list-in-voicemode-skill'
• Recent commits:
• 0855fbd [other] install.sh: install + configure mlx-audio on Apple Silicon (VM-1330)
• 7646167 [other] install.sh: offer mlx-audio on Apple Silicon even if whisper+Kokoro present (VM-1330)
• bb1b666 [other] install.sh: status-first voice-services flow, skip prompt when satisfied, reworded prompt, de-emoji banner (VM-1330)
• 073fada [other] .mcp.json: invoke launcher as bare command, not uv run (VM-1340)
• bb0f68d [other] Merge branch 'research/VM-1340-resolve-plugin-mcpjson-launcher-invocation-bare'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: @upstash/context7-pi@0.1.0 (was ctx7@0.4.2)
• Release notes:

### Minor Changes

-   f91b40c: Initial release. Adds an official Context7 extension for the [pi coding agent](https://pi.dev) — registers `resolve-library-id` and `query-docs` tools, ships the `context7-docs` skill, and exposes a `/c7-docs` slash command. Wire format, error messages, and tool descriptions are copied verbatim from `@upstash/context7-mcp` so pi and MCP clients give the LLM identical instructions and output. Self-contained — no Context7 runtime dependencies. Works out of the box at IP-based rate limits; set `CONTEXT7_API_KEY` for the higher tier. Install with `pi install npm:@upstash/context7-pi`.

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 2 docs · 4 feat · 1 fix · 2 other
• Files touched: code=37 docs=12 ci=0 test=3 other=2
• 12 new commit(s) since last check
• Latest: 7e956e5 — chore(release): version packages (#2675)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• New release: v1.5.0rc3 (was v1.4.16)
• Release notes:

## Major Improvements

*  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  The file processing pipeline supports selecting the content parsing engine and text chunking strategy either based on file extensions or on a per-file basis. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Add optional **JSON-formatted LLM** output to enhance stability in the entity and relation extraction pipeline.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

<img width="945" height="679" alt="image" src="https://github.com/user-attachments/assets/b5307383-ac00-4f98-9821-061b6e0cee00" />

## What's Broken

* **The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.

• Impact: 🟡 review (code touched)
• Commit mix: 7 build · 2 docs · 3 feat · 13 fix · 214 other · 7 perf · 3 refactor · 1 test
• Files touched: code=0 docs=25 ci=11 test=127 other=138
• 892 new commit(s) since last check
• Latest: e8fe9c1 — Merge pull request #3168 from HKUDS/feat/opensearch-batch-op
• Recent commits:
• 4c4011e [fix] fix: skip comment step on fork PRs to avoid 403
• 6d0c836 [fix] fix: pass split_by_character via chunk_options to apipeline_enqueue_documents
• 2c568e3 [build] build(deps): bump react-router-dom in /lightrag_webui in the react group
• fb6e006 [build] build(deps): bump the ui-components group
• b0f6721 [other] 📝 docs(env.example): update example configuration and documentation

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 10 other
• Files touched: code=115 docs=70 ci=0 test=65 other=93
• 33 new commit(s) since last check
• Latest: 1801156 — Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 32 docs · 8 feat · 19 fix · 9 other · 6 test
• Files touched: code=0 docs=88 ci=5 test=28 other=61
• 85 new commit(s) since last check
• Latest: 64cd1ba — fix: surface warn-only PreToolUse hooks (#2084)
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.56.0 (was v2.53.2)
• Release notes:

# Release v2.56.0

Generating release notes from v2.55.0 to HEAD
### 🔧 Maintenance

- Bump version to 2.56.0 and add changelog entry for additionalTools hook (a0d1e08)

### 📝 Other Changes

- refactor(mcp): final cleanups from review on additionalTools hook (afca30d)
- fix(mcp): security and consistency hardening for additionalTools (7699b92)
- Address review feedback: type fix, telemetry docs, new tests, dist cleanup (#801) (789ff80)
- Align additionalTools execution with core `tools/call` semantics (#800) (76228db)
- Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en) (906d6bf)
- Initial plan (afdad8b)

---

**Release Statistics:**
- 7 commits

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 1 ci · 1 feat · 1 fix · 5 other · 1 refactor
• Files touched: code=9 docs=39 ci=2 test=2 other=39
• 11 new commit(s) since last check
• Latest: 0f3d3f5 — Merge pull request #799 from czlonkowski/copilot/add-additional-tools-hook
• Recent commits:
• f1edbc9 [feat] feat(mcp): serve n8n-skills markdown via MCP Resources (#793)
• dc272cf [ci] ci(release): switch npm publishing to Trusted Publishers (OIDC) (#796)
• 41e71f8 [chore] chore: update n8n to 2.21.7 and bump version to 2.55.0 (#797)
• afdad8b [other] Initial plan
• 906d6bf [other] Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en)

claude-mem

• New release: v13.4.0 (was v13.2.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 5 docs · 8 feat · 15 fix · 2 other · 2 refactor · 1 test
• Files touched: code=102 docs=13 ci=3 test=72 other=22
• 35 new commit(s) since last check
• Latest: 8463689 — Merge pull request #2713 from thedotmack/careful-softball
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 2 feat · 3 fix · 19 other · 2 refactor · 3 test
• Files touched: code=0 docs=7 ci=1 test=5 other=17
• 37 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• 0855fbd [other] install.sh: install + configure mlx-audio on Apple Silicon (VM-1330)
• 7646167 [other] install.sh: offer mlx-audio on Apple Silicon even if whisper+Kokoro present (VM-1330)
• bb1b666 [other] install.sh: status-first voice-services flow, skip prompt when satisfied, reworded prompt, de-emoji banner (VM-1330)
• 073fada [other] .mcp.json: invoke launcher as bare command, not uv run (VM-1340)
• bb0f68d [other] Merge branch 'research/VM-1340-resolve-plugin-mcpjson-launcher-invocation-bare'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: @upstash/context7-pi@0.1.0 (was ctx7@0.4.2)
• Release notes:

### Minor Changes

-   f91b40c: Initial release. Adds an official Context7 extension for the [pi coding agent](https://pi.dev) — registers `resolve-library-id` and `query-docs` tools, ships the `context7-docs` skill, and exposes a `/c7-docs` slash command. Wire format, error messages, and tool descriptions are copied verbatim from `@upstash/context7-mcp` so pi and MCP clients give the LLM identical instructions and output. Self-contained — no Context7 runtime dependencies. Works out of the box at IP-based rate limits; set `CONTEXT7_API_KEY` for the higher tier. Install with `pi install npm:@upstash/context7-pi`.

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 2 docs · 4 feat · 1 fix · 2 other
• Files touched: code=37 docs=12 ci=0 test=3 other=2
• 12 new commit(s) since last check
• Latest: 7e956e5 — chore(release): version packages (#2675)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• New release: v1.5.0rc3 (was v1.4.16)
• Release notes:

## Major Improvements

*  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  The file processing pipeline supports selecting the content parsing engine and text chunking strategy either based on file extensions or on a per-file basis. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Add optional **JSON-formatted LLM** output to enhance stability in the entity and relation extraction pipeline.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

<img width="945" height="679" alt="image" src="https://github.com/user-attachments/assets/b5307383-ac00-4f98-9821-061b6e0cee00" />

## What's Broken

* **The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.

• Impact: 🟡 review (code touched)
• Commit mix: 5 build · 2 docs · 3 feat · 11 fix · 218 other · 7 perf · 3 refactor · 1 test
• Files touched: code=0 docs=25 ci=11 test=127 other=138
• 906 new commit(s) since last check
• Latest: 2af88d1 — Merge pull request #3170 from HKUDS/feat/postgres-graph-chunk-tx
• Recent commits:
• 12f3f01 [other] Merge pull request #3117 from HKUDS/claude/ecstatic-dijkstra-7AANg
• b246da5 [other] Merge pull request #3109 from HKUDS/dependabot/bun/lightrag_webui/react-de6a8fc1dd
• c29014e [other] Merge pull request #3110 from HKUDS/dependabot/bun/lightrag_webui/ui-components-573d650fe8
• a5c52f1 [other] ✨ feat(pipeline): surface parse/analyze progress to pipeline_status
• f0c7107 [other] Merge pull request #3118 from danielaskdd/feat/pipeline-status-parse-analyze

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 10 other
• Files touched: code=115 docs=70 ci=0 test=65 other=93
• 33 new commit(s) since last check
• Latest: 1801156 — Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 32 docs · 8 feat · 19 fix · 9 other · 6 test
• Files touched: code=0 docs=88 ci=5 test=28 other=61
• 85 new commit(s) since last check
• Latest: 64cd1ba — fix: surface warn-only PreToolUse hooks (#2084)
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.56.0 (was v2.53.2)
• Release notes:

# Release v2.56.0

Generating release notes from v2.55.0 to HEAD
### 🔧 Maintenance

- Bump version to 2.56.0 and add changelog entry for additionalTools hook (a0d1e08)

### 📝 Other Changes

- refactor(mcp): final cleanups from review on additionalTools hook (afca30d)
- fix(mcp): security and consistency hardening for additionalTools (7699b92)
- Address review feedback: type fix, telemetry docs, new tests, dist cleanup (#801) (789ff80)
- Align additionalTools execution with core `tools/call` semantics (#800) (76228db)
- Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en) (906d6bf)
- Initial plan (afdad8b)

---

**Release Statistics:**
- 7 commits

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 1 ci · 1 feat · 1 fix · 5 other · 1 refactor
• Files touched: code=9 docs=39 ci=2 test=2 other=39
• 11 new commit(s) since last check
• Latest: 0f3d3f5 — Merge pull request #799 from czlonkowski/copilot/add-additional-tools-hook
• Recent commits:
• f1edbc9 [feat] feat(mcp): serve n8n-skills markdown via MCP Resources (#793)
• dc272cf [ci] ci(release): switch npm publishing to Trusted Publishers (OIDC) (#796)
• 41e71f8 [chore] chore: update n8n to 2.21.7 and bump version to 2.55.0 (#797)
• afdad8b [other] Initial plan
• 906d6bf [other] Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en)

claude-mem

• New release: v13.4.0 (was v13.2.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 5 docs · 8 feat · 15 fix · 2 other · 2 refactor · 1 test
• Files touched: code=102 docs=13 ci=3 test=72 other=22
• 35 new commit(s) since last check
• Latest: 8463689 — Merge pull request #2713 from thedotmack/careful-softball
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 2 feat · 3 fix · 19 other · 2 refactor · 3 test
• Files touched: code=0 docs=7 ci=1 test=5 other=17
• 37 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• 0855fbd [other] install.sh: install + configure mlx-audio on Apple Silicon (VM-1330)
• 7646167 [other] install.sh: offer mlx-audio on Apple Silicon even if whisper+Kokoro present (VM-1330)
• bb1b666 [other] install.sh: status-first voice-services flow, skip prompt when satisfied, reworded prompt, de-emoji banner (VM-1330)
• 073fada [other] .mcp.json: invoke launcher as bare command, not uv run (VM-1340)
• bb0f68d [other] Merge branch 'research/VM-1340-resolve-plugin-mcpjson-launcher-invocation-bare'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: @upstash/context7-mcp@3.1.0 (was ctx7@0.4.2)
• Release notes:

### Minor Changes

-   1fb2d42: Add multi-tenant Microsoft Entra ID validation for MCP tokens. The server now detects inbound Entra v2 tokens by issuer pattern, fetches per-teamspace configuration (`tenantId`, `audience`, `requiredScope`) from the Context7 app, and verifies the token against the matching tenant's JWKS, enforcing the required scope claim when configured. User resolution happens downstream in the Context7 app against a pre-provisioned user mapping table — the MCP server only validates. Per-tenant JWKS cache and a 5-minute in-memory config cache keyed by JWT audience reduce overhead under load.

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 3 docs · 5 feat · 1 fix · 2 other
• Files touched: code=39 docs=12 ci=0 test=4 other=2
• 15 new commit(s) since last check
• Latest: 29edf15 — docs(azure-apim): post-deploy updates (Settings tab, smoke test, etc.) (#2699)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• New release: v1.5.0rc3 (was v1.4.16)
• Release notes:

## Major Improvements

*  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  The file processing pipeline supports selecting the content parsing engine and text chunking strategy either based on file extensions or on a per-file basis. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Add optional **JSON-formatted LLM** output to enhance stability in the entity and relation extraction pipeline.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

<img width="945" height="679" alt="image" src="https://github.com/user-attachments/assets/b5307383-ac00-4f98-9821-061b6e0cee00" />

## What's Broken

* **The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 5 build · 2 docs · 3 feat · 10 fix · 218 other · 8 perf · 3 refactor · 1 test
• Files touched: code=0 docs=25 ci=11 test=125 other=140
• 976 new commit(s) since last check
• Latest: 1da5814 — Merge pull request #3186 from HKUDS/claude/eager-newton-lGuVO
• Recent commits:
• 255ce3f [other] ♻️ refactor(file_atomic): simplify tmp path and reap usage
• a1368b9 [other] Merge pull request #3134 from danielaskdd/fix/atomic-index-done-callback
• 909e5fa [other] ♻️ refactor(tests): mirror feature folders under tests/
• d6054c8 [other] 🐛 fix(setup): treat opensearch-dashboards as opensearch sub-service
• 077c3f5 [other] 📝 docs(AGENTS.md): document new tests/ layout for future agents

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 10 other
• Files touched: code=115 docs=70 ci=0 test=65 other=93
• 33 new commit(s) since last check
• Latest: 1801156 — Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 32 docs · 8 feat · 19 fix · 9 other · 6 test
• Files touched: code=0 docs=88 ci=5 test=28 other=61
• 85 new commit(s) since last check
• Latest: 64cd1ba — fix: surface warn-only PreToolUse hooks (#2084)
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.56.0 (was v2.53.2)
• Release notes:

# Release v2.56.0

Generating release notes from v2.55.0 to HEAD
### 🔧 Maintenance

- Bump version to 2.56.0 and add changelog entry for additionalTools hook (a0d1e08)

### 📝 Other Changes

- refactor(mcp): final cleanups from review on additionalTools hook (afca30d)
- fix(mcp): security and consistency hardening for additionalTools (7699b92)
- Address review feedback: type fix, telemetry docs, new tests, dist cleanup (#801) (789ff80)
- Align additionalTools execution with core `tools/call` semantics (#800) (76228db)
- Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en) (906d6bf)
- Initial plan (afdad8b)

---

**Release Statistics:**
- 7 commits

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 1 ci · 1 feat · 1 fix · 5 other · 1 refactor
• Files touched: code=9 docs=39 ci=2 test=2 other=39
• 11 new commit(s) since last check
• Latest: 0f3d3f5 — Merge pull request #799 from czlonkowski/copilot/add-additional-tools-hook
• Recent commits:
• f1edbc9 [feat] feat(mcp): serve n8n-skills markdown via MCP Resources (#793)
• dc272cf [ci] ci(release): switch npm publishing to Trusted Publishers (OIDC) (#796)
• 41e71f8 [chore] chore: update n8n to 2.21.7 and bump version to 2.55.0 (#797)
• afdad8b [other] Initial plan
• 906d6bf [other] Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en)

claude-mem

• New release: v13.4.0 (was v13.2.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 5 docs · 8 feat · 15 fix · 2 other · 2 refactor · 1 test
• Files touched: code=102 docs=13 ci=3 test=72 other=22
• 35 new commit(s) since last check
• Latest: 8463689 — Merge pull request #2713 from thedotmack/careful-softball
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 2 feat · 3 fix · 19 other · 2 refactor · 3 test
• Files touched: code=0 docs=7 ci=1 test=5 other=17
• 37 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• 0855fbd [other] install.sh: install + configure mlx-audio on Apple Silicon (VM-1330)
• 7646167 [other] install.sh: offer mlx-audio on Apple Silicon even if whisper+Kokoro present (VM-1330)
• bb1b666 [other] install.sh: status-first voice-services flow, skip prompt when satisfied, reworded prompt, de-emoji banner (VM-1330)
• 073fada [other] .mcp.json: invoke launcher as bare command, not uv run (VM-1340)
• bb0f68d [other] Merge branch 'research/VM-1340-resolve-plugin-mcpjson-launcher-invocation-bare'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: @upstash/context7-mcp@3.1.0 (was ctx7@0.4.2)
• Release notes:

### Minor Changes

-   1fb2d42: Add multi-tenant Microsoft Entra ID validation for MCP tokens. The server now detects inbound Entra v2 tokens by issuer pattern, fetches per-teamspace configuration (`tenantId`, `audience`, `requiredScope`) from the Context7 app, and verifies the token against the matching tenant's JWKS, enforcing the required scope claim when configured. User resolution happens downstream in the Context7 app against a pre-provisioned user mapping table — the MCP server only validates. Per-tenant JWKS cache and a 5-minute in-memory config cache keyed by JWT audience reduce overhead under load.

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 5 feat · 1 fix · 2 other
• Files touched: code=39 docs=13 ci=0 test=4 other=2
• 16 new commit(s) since last check
• Latest: eb57986 — docs: fix verification docs typo (#2697)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 4 build · 2 docs · 3 feat · 10 fix · 219 other · 8 perf · 3 refactor · 1 test
• Files touched: code=0 docs=26 ci=11 test=124 other=140
• 1005 new commit(s) since last check
• Latest: 29bde77 — Merge pull request #3190 from HKUDS/fix/docx-outline-heading-softbreak-split
• Recent commits:
• 6f83b74 [docs] docs: add PostgreSQL image build instructions
• 8d6a6e3 [other] Merge pull request #3140 from danielaskdd/update-postgres-pg18-image
• 59a9dba [feat] feat(setup): allow custom PostgreSQL credentials for Docker deployments
• 8753999 [other] Merge pull request #3141 from HKUDS/setup-postgres-custom-docker-credentials
• b313eb4 [other] 📝 docs(ProgramingWithCore): expand addon_params documentation

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 12 other
• Files touched: code=115 docs=71 ci=0 test=65 other=93
• 35 new commit(s) since last check
• Latest: 4ec6f84 — Merge pull request #272 from vercel-labs/writing-guidelines
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 34 docs · 8 feat · 19 fix · 9 other · 6 test
• Files touched: code=0 docs=90 ci=5 test=29 other=63
• 87 new commit(s) since last check
• Latest: 99baa82 — docs: define ECC platform value loop (#2119)
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.56.1 (was v2.53.2)
• Release notes:

# Release v2.56.1

Generating release notes from v2.56.0 to HEAD
### 📝 Other Changes

- fix(versions): scope workflow_versions per instance (ced57ab)

---

**Release Statistics:**
- 1 commit
- 1 contributor
- Contributors: czlonkowski

---

## Installation

### NPM Package
```bash

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 1 ci · 1 feat · 2 fix · 6 other · 1 refactor
• Files touched: code=19 docs=39 ci=2 test=7 other=39
• 13 new commit(s) since last check
• Latest: ed0447d — Merge commit from fork
• Recent commits:
• f1edbc9 [feat] feat(mcp): serve n8n-skills markdown via MCP Resources (#793)
• dc272cf [ci] ci(release): switch npm publishing to Trusted Publishers (OIDC) (#796)
• 41e71f8 [chore] chore: update n8n to 2.21.7 and bump version to 2.55.0 (#797)
• afdad8b [other] Initial plan
• 906d6bf [other] Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en)

claude-mem

• New release: v13.4.0 (was v13.2.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 5 docs · 8 feat · 15 fix · 2 other · 2 refactor · 1 test
• Files touched: code=102 docs=13 ci=3 test=72 other=22
• 35 new commit(s) since last check
• Latest: 8463689 — Merge pull request #2713 from thedotmack/careful-softball
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 2 feat · 3 fix · 19 other · 2 refactor · 3 test
• Files touched: code=0 docs=7 ci=1 test=5 other=17
• 37 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• 0855fbd [other] install.sh: install + configure mlx-audio on Apple Silicon (VM-1330)
• 7646167 [other] install.sh: offer mlx-audio on Apple Silicon even if whisper+Kokoro present (VM-1330)
• bb1b666 [other] install.sh: status-first voice-services flow, skip prompt when satisfied, reworded prompt, de-emoji banner (VM-1330)
• 073fada [other] .mcp.json: invoke launcher as bare command, not uv run (VM-1340)
• bb0f68d [other] Merge branch 'research/VM-1340-resolve-plugin-mcpjson-launcher-invocation-bare'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.4.5 (was ctx7@0.4.2)
• Release notes:

### Patch Changes

-   2affada: `ctx7 setup` now properly supports `--antigravity`, installing skills to `.agent/skills`, a `GEMINI.md` rule section (Antigravity reads Gemini-family config), and MCP config to Antigravity 2.0's documented global path `~/.gemini/config/mcp_config.json` (with `httpUrl` for HTTP, matching the Gemini convention). Antigravity has no documented project-level MCP file, so `setup --antigravity --project --mcp` writes to the global location. Also removes the `--universal` flag from `setup`, which was advertised but silently ignored — it never propagated through agent selection, so passing it (e.g. `setup --cli --universal --project`) caused setup to fall back to auto-detection and write to the wrong directory.
-   268f52f: `ctx7 setup --api-key <KEY>` (without `--cli`, `--mcp`, or `-y`) now prompts to choose between MCP server and CLI + Skills modes. Previously, passing `--api-key` short-circuited to MCP, locking users out of the CLI + Skills option even though that mode also accepts an API key. Explicit `--mcp` / `--cli` / `--stdio` / `--oauth` / `-y` still skip the prompt as before.
-   2e97dae: Add deprecation warning to skill commands

• Impact: 🟡 review (code touched)
• Commit mix: 5 chore · 4 docs · 5 feat · 3 fix · 2 other · 1 refactor
• Files touched: code=45 docs=16 ci=0 test=4 other=2
• 20 new commit(s) since last check
• Latest: c436700 — chore(release): version packages (#2712)
• Recent commits:
• 6f0f8b7 [docs] docs(enterprise): Azure APIM deployment guide (#2636)
• 438b1df [other] update open api docs (#2637)
• dea0e43 [fix] fix(cli): declare @inquirer/core as direct dependency (#2653)
• 34fda7d [feat] feat(mcp): prompt anonymous users to sign in (#2604)
• 1ed71d7 [chore] chore(release): version packages (#2605)

lightrag

• New release: v1.5.0 (was v1.4.16)
• Release notes:

## Major Improvements

*  🎉✨**Feacture**:  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  ✨🎉**Feacture**: **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  💡✨**Feacture**: LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  💡🧠**Feacture**: **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  💡🎯**Feacture**: The file processing pipeline supports **selecting the file parsing engine and text chunking strategy either based on file extensions or on a per-file basis**. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  🚀⚡**Performance**: Optimized the vector storage persistence logic by deferring vector computations until the end of each file processing batch, enabling centralized bulk computation. **This significantly reduced the number of vector model invocations and significantly enhance the upsert speed of all vector DB LightRAG supported**.
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  🎯✨🎉**Feacture**: Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

## What's Broken

* 💥Warning: Upgrades require careful planning for users currently utilizing OpenSearch and MongoDB graph storage, **as data migration and system restarts may cause extended service outages**.
* ⚠️**The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.
* 💔**Removed deprecated field**: Deleted `QueryParam.model_func` from `lightrag/base.py`
* For OpenSearch versions prior to 3.3.0, upgrade OpenSearch before upgrading LightRAG (see https://github.com/HKUDS/LightRAG/pull/2991).

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 4 build · 1 docs · 1 feat · 8 fix · 224 other · 8 perf · 3 refactor · 1 test
• Files touched: code=0 docs=26 ci=11 test=124 other=140
• 1014 new commit(s) since last check
• Latest: c67f055 — Bump API vesion to 0305
• Recent commits:
• 6288419 [fix] fix(api): drop explicit nulls before merging chunking params
• cfd3caf [fix] fix(api): reject invalid semantic sentence_split_regex synchronously
• 20ba8a8 [fix] fix(api): skip overlap<size check for delimiter-only fixed chunking
• 8332317 [test] test(api): lock int-accepted / str-bool-rejected for semantic threshold amount
• 778ad38 [other] Merge pull request #3142 from HKUDS/feat/documents-text-chunking-api

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 2 docs · 2 feat · 19 fix · 12 other
• Files touched: code=115 docs=71 ci=0 test=65 other=93
• 35 new commit(s) since last check
• Latest: 4ec6f84 — Merge pull request #272 from vercel-labs/writing-guidelines
• Recent commits:
• 39c80b2 [feat] feat: add vercel optimize skill
• 62c3418 [fix] fix: surface ready optimize recommendations
• c5756a4 [fix] fix: redact CLI auth tokens
• 8d9ffc4 [fix] fix: preflight Observability Plus access
• 3b7d758 [fix] fix: stop fast when Observability Plus is unavailable

everything-claude-code

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 11 chore · 34 docs · 8 feat · 19 fix · 9 other · 6 test
• Files touched: code=0 docs=90 ci=5 test=29 other=63
• 87 new commit(s) since last check
• Latest: 99baa82 — docs: define ECC platform value loop (#2119)
• Recent commits:
• 6c0fbfb [docs] docs: add release plugin publication checklist
• 680aeff [test] test: enforce release publication checklist in readiness gates
• 9abe721 [docs] docs: refresh release readiness evidence
• 2ba0c62 [docs] docs: mirror agentshield fleet ticket evidence
• 99e01de [docs] docs: refresh operator dashboard evidence

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.57.0 (was v2.53.2)
• Release notes:

# Release v2.57.0

Generating release notes from v2.56.1 to HEAD
### 🔧 Maintenance

- Update n8n to 2.23.0 and bump version to 2.57.0 (2d70cce)

---

**Release Statistics:**
- 1 commit
- 1 contributor
- Contributors: czlonkowski

---

## Installation

### NPM Package
```bash

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 1 ci · 1 feat · 2 fix · 7 other · 1 refactor
• Files touched: code=19 docs=39 ci=2 test=7 other=64
• 15 new commit(s) since last check
• Latest: 56d3a29 — Merge pull request #815 from czlonkowski/update/n8n-2.23.0
• Recent commits:
• f1edbc9 [feat] feat(mcp): serve n8n-skills markdown via MCP Resources (#793)
• dc272cf [ci] ci(release): switch npm publishing to Trusted Publishers (OIDC) (#796)
• 41e71f8 [chore] chore: update n8n to 2.21.7 and bump version to 2.55.0 (#797)
• afdad8b [other] Initial plan
• 906d6bf [other] Add additionalTools engine hook (Concieved by Romuald Członkowski - www.aiadvisors.pl/en)

claude-mem

• New release: v13.4.0 (was v13.2.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 5 docs · 8 feat · 15 fix · 2 other · 2 refactor · 1 test
• Files touched: code=102 docs=13 ci=3 test=72 other=22
• 35 new commit(s) since last check
• Latest: 8463689 — Merge pull request #2713 from thedotmack/careful-softball
• Recent commits:
• bd96308 [feat] feat(skills): add design-is for Dieter Rams design audits (#2483)
• 09e74bb [feat] feat(skills): add weekly-digests skill for serial timeline narrative (#2399)
• ea057fe [feat] feat(skills): add oh-my-issues for root-cause issue clustering (#2409)
• e53d153 [fix] fix(mcp): drop root .mcp.json so plugin's mcp-search isn't duplicated (#2411)
• b0e896e [fix] fix: stop codex transcript replay after hooks migration (#2365)

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 4 docs · 2 feat · 3 fix · 19 other · 2 refactor · 3 test
• Files touched: code=0 docs=7 ci=1 test=5 other=17
• 37 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• 0855fbd [other] install.sh: install + configure mlx-audio on Apple Silicon (VM-1330)
• 7646167 [other] install.sh: offer mlx-audio on Apple Silicon even if whisper+Kokoro present (VM-1330)
• bb1b666 [other] install.sh: status-first voice-services flow, skip prompt when satisfied, reworded prompt, de-emoji banner (VM-1330)
• 073fada [other] .mcp.json: invoke launcher as bare command, not uv run (VM-1340)
• bb0f68d [other] Merge branch 'research/VM-1340-resolve-plugin-mcpjson-launcher-invocation-bare'

caveman

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 2 docs
• Files touched: code=0 docs=5 ci=0 test=0 other=0
• 2 new commit(s) since last check
• Latest: 655b7d9 — docs: feature caveman-code with a callout card after Before/After
• Recent commits:
• 2422bbe [docs] docs: add caveman-code promo banner + dancing-rock branding
• 655b7d9 [docs] docs: feature caveman-code with a callout card after Before/After

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.4.5 (was ctx7@0.4.4)
• Release notes:

### Patch Changes

-   2affada: `ctx7 setup` now properly supports `--antigravity`, installing skills to `.agent/skills`, a `GEMINI.md` rule section (Antigravity reads Gemini-family config), and MCP config to Antigravity 2.0's documented global path `~/.gemini/config/mcp_config.json` (with `httpUrl` for HTTP, matching the Gemini convention). Antigravity has no documented project-level MCP file, so `setup --antigravity --project --mcp` writes to the global location. Also removes the `--universal` flag from `setup`, which was advertised but silently ignored — it never propagated through agent selection, so passing it (e.g. `setup --cli --universal --project`) caused setup to fall back to auto-detection and write to the wrong directory.
-   268f52f: `ctx7 setup --api-key <KEY>` (without `--cli`, `--mcp`, or `-y`) now prompts to choose between MCP server and CLI + Skills modes. Previously, passing `--api-key` short-circuited to MCP, locking users out of the CLI + Skills option even though that mode also accepts an API key. Explicit `--mcp` / `--cli` / `--stdio` / `--oauth` / `-y` still skip the prompt as before.
-   2e97dae: Add deprecation warning to skill commands

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 3 docs · 1 feat · 2 fix · 1 refactor
• Files touched: code=19 docs=11 ci=0 test=2 other=0
• 10 new commit(s) since last check
• Latest: 93af8bf — docs: mention Node requirement for ctx7 setup (#2640)
• Recent commits:
• 7e956e5 [chore] chore(release): version packages (#2675)
• 1fb2d42 [feat] feat(mcp): multi-tenant Entra ID validation (#2629)
• 801e112 [chore] chore(release): version packages (#2698)
• 29edf15 [docs] docs(azure-apim): post-deploy updates (Settings tab, smoke test, etc.) (#2699)
• eb57986 [docs] docs: fix verification docs typo (#2697)

lightrag

• New release: v1.5.0 (was v1.5.0rc2)
• Release notes:

## Major Improvements

*  🎉✨**Feacture**:  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  ✨🎉**Feacture**: **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  💡✨**Feacture**: LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  💡🧠**Feacture**: **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  💡🎯**Feacture**: The file processing pipeline supports **selecting the file parsing engine and text chunking strategy either based on file extensions or on a per-file basis**. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  🚀⚡**Performance**: Optimized the vector storage persistence logic by deferring vector computations until the end of each file processing batch, enabling centralized bulk computation. **This significantly reduced the number of vector model invocations and significantly enhance the upsert speed of all vector DB LightRAG supported**.
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  🎯✨🎉**Feacture**: Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

## What's Broken

* 💥Warning: Upgrades require careful planning for users currently utilizing OpenSearch and MongoDB graph storage, **as data migration and system restarts may cause extended service outages**.
* ⚠️**The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.
* 💔**Removed deprecated field**: Deleted `QueryParam.model_func` from `lightrag/base.py`
* For OpenSearch versions prior to 3.3.0, upgrade OpenSearch before upgrading LightRAG (see https://github.com/HKUDS/LightRAG/pull/2991).

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 5 build · 1 docs · 1 feat · 5 fix · 226 other · 8 perf · 3 refactor · 1 test
• Files touched: code=0 docs=14 ci=4 test=178 other=72
• 298 new commit(s) since last check
• Latest: 4396da4 — 🔧 chore(uv.lock): widen redis version range
• Recent commits:
• 8332317 [test] test(api): lock int-accepted / str-bool-rejected for semantic threshold amount
• 778ad38 [other] Merge pull request #3142 from HKUDS/feat/documents-text-chunking-api
• 551f006 [perf] perf(opensearch): defer vector embedding until flush
• 74181d1 [other] 🐛 fix(opensearch): preserve pending vector docs
• a9884a2 [other] 🐛 fix(opensearch): harden embedding flush consistency

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 4 other
• Files touched: code=5 docs=6 ci=0 test=5 other=12
• 4 new commit(s) since last check
• Latest: 4ec6f84 — Merge pull request #272 from vercel-labs/writing-guidelines
• Recent commits:
• 2f423a2 [other] Add writing-guidelines skill
• 583754a [other] Harden vercel optimize report safety
• 1801156 [other] Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• 4ec6f84 [other] Merge pull request #272 from vercel-labs/writing-guidelines

everything-claude-code

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 5 docs · 7 feat · 8 fix · 1 other
• Files touched: code=0 docs=70 ci=2 test=13 other=37
• 24 new commit(s) since last check
• Latest: 0f84c0e — feat: add ECC2 local control pane (#2131)
• Recent commits:
• 5bacdf4 [feat] feat: publish ECC 2.0 skill pack surfaces
• 870c5eb [chore] chore(deps): bump actions/stale to 10.3.0 (#2045)
• dcee223 [fix] fix(docs): update Chinese marketplace URL (#2050)
• 7485e41 [docs] docs(security-guide): add LLM Safe Haven reference (#2034)
• 3ffab63 [fix] fix(nextjs-turbopack): document proxy.ts middleware filename (#2033)

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.57.1 (was v2.56.0)
• Release notes:

# Release v2.57.1

Generating release notes from v2.57.0 to HEAD
### 📝 Other Changes

- fix(credentials): harden list input validation and strip data field (#816) (a5d56fa)
- fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816) (ca45d9d)

---

**Release Statistics:**
- 2 commits
- 1 contributor
- Contributors: czlonkowski

---

## Installation

### NPM Package

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 3 fix · 3 other
• Files touched: code=13 docs=2 ci=0 test=7 other=35
• 7 new commit(s) since last check
• Latest: 9e30942 — Merge pull request #830 from czlonkowski/fix/credential-pagination-816
• Recent commits:
• ced57ab [fix] fix(versions): scope workflow_versions per instance
• ed0447d [other] Merge commit from fork
• 2d70cce [chore] chore: update n8n to 2.23.0 and bump version to 2.57.0
• 56d3a29 [other] Merge pull request #815 from czlonkowski/update/n8n-2.23.0
• ca45d9d [fix] fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816)

claude-mem

• New release: v13.4.0 (was v13.3.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 4 docs · 5 feat · 13 fix · 2 other · 2 refactor · 1 test
• Files touched: code=100 docs=9 ci=3 test=69 other=20
• 28 new commit(s) since last check
• Latest: 8463689 — Merge pull request #2713 from thedotmack/careful-softball
• Recent commits:
• c46563c [docs] docs: add plan-07..11 architectural fix plans
• 61fe70a [refactor] refactor: remove retry/observation queue, replace with SessionMessageBuffer
• 7a6eaca [fix] fix(build): plan-10 artifact hygiene — green typecheck, drop better-auth from worker, CI gate
• a8cf671 [fix] fix(tests+pipeline): repair test suite (46->7 fails) + plan-09 schema self-healing
• 90f7049 [fix] fix(opencode+openclaw): plan-08 event-contract correctness — capture actually works

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 2 docs · 2 feat · 2 fix · 11 other
• Files touched: code=0 docs=3 ci=1 test=3 other=15
• 21 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• c183c91 [other] deps(deps): bump starlette from 0.47.3 to 0.49.1
• a3608c6 [other] Merge pull request #438 from mbailey/dependabot/uv/starlette-0.49.1
• 276f1be [chore] chore(dependabot): group every bump (major/minor/patch) into one weekly PR
• d0a7629 [other] Merge branch 'chore/dependabot-group-all-bumps'
• af7232b [fix] fix(VM-1398): systemd Restart=always so kokoro survives UVICORN_LIMIT_MAX_REQUESTS

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.5.1 (was ctx7@0.4.4)
• Release notes:

### Patch Changes

- ea91d7d: `ctx7 login` now always uses the device-code flow. The localhost-callback path is removed — every install (laptop, SSH, Codespace, Docker, CI) goes through the same boxed prompt and verification page. Drops the `--device` flag (it was the opt-in for what's now the default). Older CLI versions (≤ 0.5.0) continue to work against the unchanged auth endpoints, so pinned installs are unaffected.

• Impact: 🟡 review (code touched)
• Commit mix: 5 chore · 3 docs · 3 feat · 2 fix · 1 refactor
• Files touched: code=23 docs=11 ci=0 test=4 other=1
• 14 new commit(s) since last check
• Latest: d82bcc9 — chore(release): version packages (#2721)
• Recent commits:
• 7e956e5 [chore] chore(release): version packages (#2675)
• 1fb2d42 [feat] feat(mcp): multi-tenant Entra ID validation (#2629)
• 801e112 [chore] chore(release): version packages (#2698)
• 29edf15 [docs] docs(azure-apim): post-deploy updates (Settings tab, smoke test, etc.) (#2699)
• eb57986 [docs] docs: fix verification docs typo (#2697)

lightrag

• New release: v1.5.0 (was v1.5.0rc2)
• Release notes:

## Major Improvements

*  🎉✨**Feacture**:  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  ✨🎉**Feacture**: **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  💡✨**Feacture**: LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  💡🧠**Feacture**: **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  💡🎯**Feacture**: The file processing pipeline supports **selecting the file parsing engine and text chunking strategy either based on file extensions or on a per-file basis**. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  🚀⚡**Performance**: Optimized the vector storage persistence logic by deferring vector computations until the end of each file processing batch, enabling centralized bulk computation. **This significantly reduced the number of vector model invocations and significantly enhance the upsert speed of all vector DB LightRAG supported**.
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  🎯✨🎉**Feacture**: Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

## What's Broken

* 💥Warning: Upgrades require careful planning for users currently utilizing OpenSearch and MongoDB graph storage, **as data migration and system restarts may cause extended service outages**.
* ⚠️**The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.
* 💔**Removed deprecated field**: Deleted `QueryParam.model_func` from `lightrag/base.py`
* For OpenSearch versions prior to 3.3.0, upgrade OpenSearch before upgrading LightRAG (see https://github.com/HKUDS/LightRAG/pull/2991).

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 10 build · 1 docs · 1 feat · 5 fix · 225 other · 5 perf · 3 refactor
• Files touched: code=0 docs=18 ci=4 test=181 other=73
• 320 new commit(s) since last check
• Latest: aa42727 — Merge pull request #3206 from he-yufeng/fix/postgres-search-path-table-check
• Recent commits:
• 1399a66 [other] 🐛 fix(operate): correct final_data key path in debug log
• cb80c5b [other] Merge pull request #3148 from danielaskdd/fix/build-context-log-counts
• 6fb444f [perf] perf(milvus): defer vector embedding until flush
• d411d7c [perf] perf(mongo): defer vector embedding until flush
• 85cc7db [perf] perf(qdrant): defer vector embedding until flush

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 4 other
• Files touched: code=5 docs=6 ci=0 test=5 other=12
• 4 new commit(s) since last check
• Latest: 4ec6f84 — Merge pull request #272 from vercel-labs/writing-guidelines
• Recent commits:
• 2f423a2 [other] Add writing-guidelines skill
• 583754a [other] Harden vercel optimize report safety
• 1801156 [other] Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• 4ec6f84 [other] Merge pull request #272 from vercel-labs/writing-guidelines

everything-claude-code

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 5 docs · 8 feat · 8 fix · 1 other
• Files touched: code=0 docs=73 ci=2 test=14 other=37
• 25 new commit(s) since last check
• Latest: bc8e12b — feat: add dynamic workflow team orchestration surface
• Recent commits:
• 5bacdf4 [feat] feat: publish ECC 2.0 skill pack surfaces
• 870c5eb [chore] chore(deps): bump actions/stale to 10.3.0 (#2045)
• dcee223 [fix] fix(docs): update Chinese marketplace URL (#2050)
• 7485e41 [docs] docs(security-guide): add LLM Safe Haven reference (#2034)
• 3ffab63 [fix] fix(nextjs-turbopack): document proxy.ts middleware filename (#2033)

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.57.1 (was v2.56.0)
• Release notes:

# Release v2.57.1

Generating release notes from v2.57.0 to HEAD
### 📝 Other Changes

- fix(credentials): harden list input validation and strip data field (#816) (a5d56fa)
- fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816) (ca45d9d)

---

**Release Statistics:**
- 2 commits
- 1 contributor
- Contributors: czlonkowski

---

## Installation

### NPM Package

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 3 fix · 3 other
• Files touched: code=13 docs=2 ci=0 test=7 other=35
• 7 new commit(s) since last check
• Latest: 9e30942 — Merge pull request #830 from czlonkowski/fix/credential-pagination-816
• Recent commits:
• ced57ab [fix] fix(versions): scope workflow_versions per instance
• ed0447d [other] Merge commit from fork
• 2d70cce [chore] chore: update n8n to 2.23.0 and bump version to 2.57.0
• 56d3a29 [other] Merge pull request #815 from czlonkowski/update/n8n-2.23.0
• ca45d9d [fix] fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816)

claude-mem

• New release: v13.4.0 (was v13.3.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 4 docs · 5 feat · 13 fix · 2 other · 2 refactor · 1 test
• Files touched: code=100 docs=9 ci=3 test=69 other=20
• 28 new commit(s) since last check
• Latest: 8463689 — Merge pull request #2713 from thedotmack/careful-softball
• Recent commits:
• c46563c [docs] docs: add plan-07..11 architectural fix plans
• 61fe70a [refactor] refactor: remove retry/observation queue, replace with SessionMessageBuffer
• 7a6eaca [fix] fix(build): plan-10 artifact hygiene — green typecheck, drop better-auth from worker, CI gate
• a8cf671 [fix] fix(tests+pipeline): repair test suite (46->7 fails) + plan-09 schema self-healing
• 90f7049 [fix] fix(opencode+openclaw): plan-08 event-contract correctness — capture actually works

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 2 docs · 2 feat · 2 fix · 11 other
• Files touched: code=0 docs=3 ci=1 test=3 other=15
• 21 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• c183c91 [other] deps(deps): bump starlette from 0.47.3 to 0.49.1
• a3608c6 [other] Merge pull request #438 from mbailey/dependabot/uv/starlette-0.49.1
• 276f1be [chore] chore(dependabot): group every bump (major/minor/patch) into one weekly PR
• d0a7629 [other] Merge branch 'chore/dependabot-group-all-bumps'
• af7232b [fix] fix(VM-1398): systemd Restart=always so kokoro survives UVICORN_LIMIT_MAX_REQUESTS

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.5.1 (was ctx7@0.4.4)
• Release notes:

### Patch Changes

- ea91d7d: `ctx7 login` now always uses the device-code flow. The localhost-callback path is removed — every install (laptop, SSH, Codespace, Docker, CI) goes through the same boxed prompt and verification page. Drops the `--device` flag (it was the opt-in for what's now the default). Older CLI versions (≤ 0.5.0) continue to work against the unchanged auth endpoints, so pinned installs are unaffected.

• Impact: 🟡 review (code touched)
• Commit mix: 5 chore · 4 docs · 3 feat · 2 fix · 1 other · 1 refactor
• Files touched: code=23 docs=13 ci=0 test=4 other=1
• 16 new commit(s) since last check
• Latest: 7dde338 — docs: add Policies and Rules How To pages (#2724)
• Recent commits:
• 7e956e5 [chore] chore(release): version packages (#2675)
• 1fb2d42 [feat] feat(mcp): multi-tenant Entra ID validation (#2629)
• 801e112 [chore] chore(release): version packages (#2698)
• 29edf15 [docs] docs(azure-apim): post-deploy updates (Settings tab, smoke test, etc.) (#2699)
• eb57986 [docs] docs: fix verification docs typo (#2697)

lightrag

• New release: v1.5.0 (was v1.5.0rc2)
• Release notes:

## Major Improvements

*  🎉✨**Feacture**:  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  ✨🎉**Feacture**: **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  💡✨**Feacture**: LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  💡🧠**Feacture**: **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  💡🎯**Feacture**: The file processing pipeline supports **selecting the file parsing engine and text chunking strategy either based on file extensions or on a per-file basis**. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  🚀⚡**Performance**: Optimized the vector storage persistence logic by deferring vector computations until the end of each file processing batch, enabling centralized bulk computation. **This significantly reduced the number of vector model invocations and significantly enhance the upsert speed of all vector DB LightRAG supported**.
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  🎯✨🎉**Feacture**: Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

## What's Broken

* 💥Warning: Upgrades require careful planning for users currently utilizing OpenSearch and MongoDB graph storage, **as data migration and system restarts may cause extended service outages**.
* ⚠️**The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.
* 💔**Removed deprecated field**: Deleted `QueryParam.model_func` from `lightrag/base.py`
* For OpenSearch versions prior to 3.3.0, upgrade OpenSearch before upgrading LightRAG (see https://github.com/HKUDS/LightRAG/pull/2991).

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 10 build · 3 feat · 6 fix · 229 other · 1 perf · 1 refactor
• Files touched: code=0 docs=18 ci=4 test=181 other=75
• 332 new commit(s) since last check
• Latest: c73c373 — 📝 docs(README): clarify configuration stages
• Recent commits:
• 2a43eb0 [other] 📝 docs(milvus): document drop buffer aliasing risk
• 02be2a2 [other] Merge pull request #3149 from HKUDS/claude/lazy-embed-milvus
• 004cfa1 [other] 📝 docs(mongo): clarify deferred flush behavior
• 46931df [other] Merge branch 'main' into claude/lazy-embed-mongo
• 3f43306 [other] 📝 docs(mongo): improve vector storage error logs

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 4 other
• Files touched: code=5 docs=6 ci=0 test=5 other=12
• 4 new commit(s) since last check
• Latest: 4ec6f84 — Merge pull request #272 from vercel-labs/writing-guidelines
• Recent commits:
• 2f423a2 [other] Add writing-guidelines skill
• 583754a [other] Harden vercel optimize report safety
• 1801156 [other] Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• 4ec6f84 [other] Merge pull request #272 from vercel-labs/writing-guidelines

everything-claude-code

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 5 docs · 10 feat · 8 fix · 1 other
• Files touched: code=0 docs=73 ci=2 test=17 other=47
• 27 new commit(s) since last check
• Latest: 7113b5b — feat: MCP inventory (ecc.mcp.v1) — unified cross-harness MCP config view (#2146)
• Recent commits:
• 5bacdf4 [feat] feat: publish ECC 2.0 skill pack surfaces
• 870c5eb [chore] chore(deps): bump actions/stale to 10.3.0 (#2045)
• dcee223 [fix] fix(docs): update Chinese marketplace URL (#2050)
• 7485e41 [docs] docs(security-guide): add LLM Safe Haven reference (#2034)
• 3ffab63 [fix] fix(nextjs-turbopack): document proxy.ts middleware filename (#2033)

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.57.1 (was v2.56.0)
• Release notes:

# Release v2.57.1

Generating release notes from v2.57.0 to HEAD
### 📝 Other Changes

- fix(credentials): harden list input validation and strip data field (#816) (a5d56fa)
- fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816) (ca45d9d)

---

**Release Statistics:**
- 2 commits
- 1 contributor
- Contributors: czlonkowski

---

## Installation

### NPM Package

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 3 fix · 3 other
• Files touched: code=13 docs=2 ci=0 test=7 other=35
• 7 new commit(s) since last check
• Latest: 9e30942 — Merge pull request #830 from czlonkowski/fix/credential-pagination-816
• Recent commits:
• ced57ab [fix] fix(versions): scope workflow_versions per instance
• ed0447d [other] Merge commit from fork
• 2d70cce [chore] chore: update n8n to 2.23.0 and bump version to 2.57.0
• 56d3a29 [other] Merge pull request #815 from czlonkowski/update/n8n-2.23.0
• ca45d9d [fix] fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816)

claude-mem

• New release: v13.4.0 (was v13.3.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 4 docs · 5 feat · 36 fix · 6 other · 2 refactor · 1 test
• Files touched: code=117 docs=23 ci=4 test=86 other=28
• 55 new commit(s) since last check
• Latest: 671de5e — Prevent observer file edits from self-recording (#2519)
• Recent commits:
• c46563c [docs] docs: add plan-07..11 architectural fix plans
• 61fe70a [refactor] refactor: remove retry/observation queue, replace with SessionMessageBuffer
• 7a6eaca [fix] fix(build): plan-10 artifact hygiene — green typecheck, drop better-auth from worker, CI gate
• a8cf671 [fix] fix(tests+pipeline): repair test suite (46->7 fails) + plan-09 schema self-healing
• 90f7049 [fix] fix(opencode+openclaw): plan-08 event-contract correctness — capture actually works

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 2 docs · 2 feat · 2 fix · 11 other
• Files touched: code=0 docs=3 ci=1 test=3 other=15
• 21 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• c183c91 [other] deps(deps): bump starlette from 0.47.3 to 0.49.1
• a3608c6 [other] Merge pull request #438 from mbailey/dependabot/uv/starlette-0.49.1
• 276f1be [chore] chore(dependabot): group every bump (major/minor/patch) into one weekly PR
• d0a7629 [other] Merge branch 'chore/dependabot-group-all-bumps'
• af7232b [fix] fix(VM-1398): systemd Restart=always so kokoro survives UVICORN_LIMIT_MAX_REQUESTS

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.5.1 (was ctx7@0.4.4)
• Release notes:

### Patch Changes

- ea91d7d: `ctx7 login` now always uses the device-code flow. The localhost-callback path is removed — every install (laptop, SSH, Codespace, Docker, CI) goes through the same boxed prompt and verification page. Drops the `--device` flag (it was the opt-in for what's now the default). Older CLI versions (≤ 0.5.0) continue to work against the unchanged auth endpoints, so pinned installs are unaffected.

• Impact: 🟡 review (code touched)
• Commit mix: 5 chore · 4 docs · 3 feat · 2 fix · 1 other · 1 refactor
• Files touched: code=23 docs=13 ci=0 test=4 other=1
• 16 new commit(s) since last check
• Latest: 7dde338 — docs: add Policies and Rules How To pages (#2724)
• Recent commits:
• 7e956e5 [chore] chore(release): version packages (#2675)
• 1fb2d42 [feat] feat(mcp): multi-tenant Entra ID validation (#2629)
• 801e112 [chore] chore(release): version packages (#2698)
• 29edf15 [docs] docs(azure-apim): post-deploy updates (Settings tab, smoke test, etc.) (#2699)
• eb57986 [docs] docs: fix verification docs typo (#2697)

lightrag

• New release: v1.5.0 (was v1.5.0rc2)
• Release notes:

## Major Improvements

*  🎉✨**Feacture**:  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  ✨🎉**Feacture**: **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  💡✨**Feacture**: LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  💡🧠**Feacture**: **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  💡🎯**Feacture**: The file processing pipeline supports **selecting the file parsing engine and text chunking strategy either based on file extensions or on a per-file basis**. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  🚀⚡**Performance**: Optimized the vector storage persistence logic by deferring vector computations until the end of each file processing batch, enabling centralized bulk computation. **This significantly reduced the number of vector model invocations and significantly enhance the upsert speed of all vector DB LightRAG supported**.
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  🎯✨🎉**Feacture**: Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

## What's Broken

* 💥Warning: Upgrades require careful planning for users currently utilizing OpenSearch and MongoDB graph storage, **as data migration and system restarts may cause extended service outages**.
* ⚠️**The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.
* 💔**Removed deprecated field**: Deleted `QueryParam.model_func` from `lightrag/base.py`
* For OpenSearch versions prior to 3.3.0, upgrade OpenSearch before upgrading LightRAG (see https://github.com/HKUDS/LightRAG/pull/2991).

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 10 build · 3 feat · 6 fix · 229 other · 1 perf · 1 refactor
• Files touched: code=0 docs=20 ci=4 test=183 other=77
• 351 new commit(s) since last check
• Latest: 3da423b — Merge branch 'claude/pr-3216-table-splitting-LNMim'
• Recent commits:
• f2fe2f5 [other] 🐛 fix(postgres_impl): drop stale embeddings from response
• c50e1df [other] 🐛 fix(postgres): re-raise delete failures
• 04fb1a2 [other] 🐛 fix(postgres): enforce lifecycle safety for vector ops
• b13e9e7 [other] 🐛 fix(postgres): log vector storage drop failures
• 5c5a7c6 [other] 🐛 fix(postgres_impl): improve vector flush error handling

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 4 other
• Files touched: code=5 docs=6 ci=0 test=5 other=12
• 4 new commit(s) since last check
• Latest: 4ec6f84 — Merge pull request #272 from vercel-labs/writing-guidelines
• Recent commits:
• 2f423a2 [other] Add writing-guidelines skill
• 583754a [other] Harden vercel optimize report safety
• 1801156 [other] Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• 4ec6f84 [other] Merge pull request #272 from vercel-labs/writing-guidelines

everything-claude-code

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 12 docs · 23 feat · 34 fix · 3 other · 2 test
• Files touched: code=0 docs=256 ci=8 test=0 other=36
• 78 new commit(s) since last check
• Latest: 90dfd95 — feat: add orch-* orchestrator skill family (#2153)
• Recent commits:
• 5bacdf4 [feat] feat: publish ECC 2.0 skill pack surfaces
• 870c5eb [chore] chore(deps): bump actions/stale to 10.3.0 (#2045)
• dcee223 [fix] fix(docs): update Chinese marketplace URL (#2050)
• 7485e41 [docs] docs(security-guide): add LLM Safe Haven reference (#2034)
• 3ffab63 [fix] fix(nextjs-turbopack): document proxy.ts middleware filename (#2033)

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.57.1 (was v2.56.0)
• Release notes:

# Release v2.57.1

Generating release notes from v2.57.0 to HEAD
### 📝 Other Changes

- fix(credentials): harden list input validation and strip data field (#816) (a5d56fa)
- fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816) (ca45d9d)

---

**Release Statistics:**
- 2 commits
- 1 contributor
- Contributors: czlonkowski

---

## Installation

### NPM Package

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 3 fix · 3 other
• Files touched: code=13 docs=2 ci=0 test=7 other=35
• 7 new commit(s) since last check
• Latest: 9e30942 — Merge pull request #830 from czlonkowski/fix/credential-pagination-816
• Recent commits:
• ced57ab [fix] fix(versions): scope workflow_versions per instance
• ed0447d [other] Merge commit from fork
• 2d70cce [chore] chore: update n8n to 2.23.0 and bump version to 2.57.0
• 56d3a29 [other] Merge pull request #815 from czlonkowski/update/n8n-2.23.0
• ca45d9d [fix] fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816)

claude-mem

• New release: v13.4.0 (was v13.3.0)
• Release notes:

## v13.4.0 — Defect backlog clearance + provider extensibility

Clears a large defect backlog (plans 01–11 plus standalone fixes) and adds provider configurability. Test suite moved 46 → 0 failing and typecheck 24 → 0 errors over the branch.

### Features
- **Configurable OpenAI-compatible base URL** for the OpenRouter provider (`CLAUDE_MEM_OPENROUTER_BASE_URL`) — point claude-mem at DeepSeek, LM Studio, or any custom OpenAI-compatible endpoint.

### Fixes (highlights)
- **Spawn contract (plan-02):** canonical `${CLAUDE_PLUGIN_ROOT}` resolution + Windows spawn fixes (codex.cmd, chroma-mcp cmd.exe quoting).
- **Worker lifecycle (plan-03):** Windows PID-reuse start-token guard.
- **Output fidelity (plan-11):** commit-hash verification before persist; null-`cwd` no longer strips every hex string from summaries.
- **SQLite self-healing:** schema repair via `sqlite3 .recover`; close DB handle on repair error paths (no leaked write lock).
- **SessionMessageBuffer:** `clear()` now also resets the dedup set, so a previously-seen toolUseId can re-enter.
- **Standalone:** project name, dot-path encoding, path-match, CLAUDE.md denylist.

### CI / tests
- New CI workflow (typecheck · build · test · bundle-size + docker pg+valkey e2e) made green; removed npm-lockfile dependency to match the repo's no-committed-lockfile convention.
- Fixed `mock.module` logger leakage across test files and guarded sqlite3 `.recover` capability so CI runs cleanly.

Full diff: https://github.com/thedotmack/claude-mem/pull/2701

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 4 docs · 5 feat · 36 fix · 6 other · 2 refactor · 1 test
• Files touched: code=117 docs=23 ci=4 test=86 other=28
• 55 new commit(s) since last check
• Latest: 671de5e — Prevent observer file edits from self-recording (#2519)
• Recent commits:
• c46563c [docs] docs: add plan-07..11 architectural fix plans
• 61fe70a [refactor] refactor: remove retry/observation queue, replace with SessionMessageBuffer
• 7a6eaca [fix] fix(build): plan-10 artifact hygiene — green typecheck, drop better-auth from worker, CI gate
• a8cf671 [fix] fix(tests+pipeline): repair test suite (46->7 fails) + plan-09 schema self-healing
• 90f7049 [fix] fix(opencode+openclaw): plan-08 event-contract correctness — capture actually works

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 2 docs · 2 feat · 2 fix · 11 other
• Files touched: code=0 docs=3 ci=1 test=3 other=15
• 21 new commit(s) since last check
• Latest: e1564f2 — Merge branch 'feat/VM-1024-focus-mode-cli-command-offgentleaggressive-modes'
• Recent commits:
• c183c91 [other] deps(deps): bump starlette from 0.47.3 to 0.49.1
• a3608c6 [other] Merge pull request #438 from mbailey/dependabot/uv/starlette-0.49.1
• 276f1be [chore] chore(dependabot): group every bump (major/minor/patch) into one weekly PR
• d0a7629 [other] Merge branch 'chore/dependabot-group-all-bumps'
• af7232b [fix] fix(VM-1398): systemd Restart=always so kokoro survives UVICORN_LIMIT_MAX_REQUESTS

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.5.1 (was ctx7@0.4.4)
• Release notes:

### Patch Changes

- ea91d7d: `ctx7 login` now always uses the device-code flow. The localhost-callback path is removed — every install (laptop, SSH, Codespace, Docker, CI) goes through the same boxed prompt and verification page. Drops the `--device` flag (it was the opt-in for what's now the default). Older CLI versions (≤ 0.5.0) continue to work against the unchanged auth endpoints, so pinned installs are unaffected.

• Impact: 🟡 review (code touched)
• Commit mix: 5 chore · 4 docs · 3 feat · 4 fix · 1 other · 1 refactor
• Files touched: code=28 docs=15 ci=0 test=6 other=1
• 18 new commit(s) since last check
• Latest: c03bc9c — fix(cli): use XDG dirs for context files (#2678)
• Recent commits:
• 7e956e5 [chore] chore(release): version packages (#2675)
• 1fb2d42 [feat] feat(mcp): multi-tenant Entra ID validation (#2629)
• 801e112 [chore] chore(release): version packages (#2698)
• 29edf15 [docs] docs(azure-apim): post-deploy updates (Settings tab, smoke test, etc.) (#2699)
• eb57986 [docs] docs: fix verification docs typo (#2697)

lightrag

• New release: v1.5.0 (was v1.5.0rc2)
• Release notes:

## Major Improvements

*  🎉✨**Feacture**:  **LightRAG now supports multimodal document processing** and can fully leverage images, tables, and formulas within documents to answer queries. **All RagAnything’s multimodal processing capabilities are merged into LightRAG**; ( [RagAnything](https://github.com/HKUDS/RAG-Anything) will no longer receive core feature updates or maintenance going forward)
*  ✨🎉**Feacture**: **Fully upgraded the document processing pipeline**, with support for using **MinerU** and **Docling** to parse and extract file content, seamlessly integrating with multimodal content analysis and entity-relationship extraction.
*  💡✨**Feacture**: LightRAG now introduces **Native Parsing**, enabling high-quality content extraction from DOCX documents. It supports accurate reconstruction of Word auto-numbering, as well as extraction of images, tables, and formulas, providing seamless integration with multimodal content analysis and entity-relationship extraction. Expanded format support for the **Native Parser** is coming soon.
*  💡🧠**Feacture**: **Introduced four selectable text chunking strategies: `Fix`, `Recursive`, `Vector`, and `Paragraph`.** The parameters for each chunking strategy can be configured through environment variables.
*  💡🎯**Feacture**: The file processing pipeline supports **selecting the file parsing engine and text chunking strategy either based on file extensions or on a per-file basis**. For detailed usage instructions, refer to [FileProcessingPipeline.md](https://github.com/HKUDS/LightRAG/blob/main/docs/FileProcessingPipeline.md).
*  🚀⚡**Performance**: Optimized the vector storage persistence logic by deferring vector computations until the end of each file processing batch, enabling centralized bulk computation. **This significantly reduced the number of vector model invocations and significantly enhance the upsert speed of all vector DB LightRAG supported**.
*  Enable **task-aware embedding** support for asymmetric models, including `voyage-3`, `text-embedding-004`, `embed-multilingual-v3.0`, and `jina-embeddings-v3`.
*  Improved entity/relation extraction reliability by **introducing LLM  JSON-structured output**; set ENTITY_EXTRACTION_USE_JSON=true to enable it.
*  Introduce `ENTITY_TYPE_PROMPT_FILE` to empower users with enhanced guidance for LLM-driven **entity type recognition** and extraction.
*  Fully support Amazon and Anthropic models on **AWS Bedrock API**.
*  🎯✨🎉**Feacture**: Implemented **role-specific LLM configuration** support, introducing four distinct roles: EXTRACT, QUERY, KEYWORDS, and VLM, each with independent LLM settings. It is recommended to configure the KEYWORDS role with a small-parameter, non-reasoning, high-speed model to optimize query latency; the EXTRACT role with a medium-parameter, non-reasoning model to balance accuracy and throughput; and the QUERY role with a large-parameter reasoning model to enhance query quality. For detailed usage instructions, refer to [RoleSpecificLLMConfiguration.md](https://github.com/HKUDS/LightRAG/blob/main/docs/RoleSpecificLLMConfiguration.md).

## What's Broken

* 💥Warning: Upgrades require careful planning for users currently utilizing OpenSearch and MongoDB graph storage, **as data migration and system restarts may cause extended service outages**.
* ⚠️**The `ENTITY_TYPES` environment variable has been deprecated**; Please replace it with ENTITY_TYPE_PROMPT_FILE before launching this new version.
* 💔**Removed deprecated field**: Deleted `QueryParam.model_func` from `lightrag/base.py`
* For OpenSearch versions prior to 3.3.0, upgrade OpenSearch before upgrading LightRAG (see https://github.com/HKUDS/LightRAG/pull/2991).

• Impact: 🔴 likely matters (breaking change)
• Commit mix: 6 build · 2 chore · 1 docs · 8 feat · 13 fix · 217 other · 1 perf · 2 refactor
• Files touched: code=0 docs=20 ci=5 test=192 other=85
• 387 new commit(s) since last check
• Latest: 2a55430 — Bump API version to 0307
• Recent commits:
• b560673 [fix] fix(sidecar): preserve source spans through hard-fallback rejoin
• 1169747 [other] 🐛 fix(sidecar): skip unlocatable replacement-char chunks
• 0dcb9a7 [other] 🐛 fix(chunker): recover source spans with forward cursor
• 5982de3 [other] 🐛 fix(chunker): preserve source spans for repeated text
• 8d94b0b [other] 🐛 fix(chunker): preserve recursive chunk source spans

vercel-agent-skills

• Impact: 🟡 review (code touched)
• Commit mix: 4 other
• Files touched: code=5 docs=6 ci=0 test=5 other=12
• 4 new commit(s) since last check
• Latest: 4ec6f84 — Merge pull request #272 from vercel-labs/writing-guidelines
• Recent commits:
• 2f423a2 [other] Add writing-guidelines skill
• 583754a [other] Harden vercel optimize report safety
• 1801156 [other] Merge pull request #273 from vercel-labs/johnphamous/vercel-optimize-observation-safety
• 4ec6f84 [other] Merge pull request #272 from vercel-labs/writing-guidelines

everything-claude-code

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 12 docs · 23 feat · 34 fix · 3 other · 2 test
• Files touched: code=0 docs=256 ci=8 test=0 other=36
• 78 new commit(s) since last check
• Latest: 90dfd95 — feat: add orch-* orchestrator skill family (#2153)
• Recent commits:
• 5bacdf4 [feat] feat: publish ECC 2.0 skill pack surfaces
• 870c5eb [chore] chore(deps): bump actions/stale to 10.3.0 (#2045)
• dcee223 [fix] fix(docs): update Chinese marketplace URL (#2050)
• 7485e41 [docs] docs(security-guide): add LLM Safe Haven reference (#2034)
• 3ffab63 [fix] fix(nextjs-turbopack): document proxy.ts middleware filename (#2033)

superpowers

• Impact: 🟢 skim (docs/ci/test only)
• Commit mix: 1 other
• Files touched: code=0 docs=5 ci=4 test=0 other=0
• 1 new commit(s) since last check
• Latest: 6fd4507 — Require contributors to disclose authoring environment and target dev
• Recent commits:
• 6fd4507 [other] Require contributors to disclose authoring environment and target dev

n8n-mcp

• New release: v2.57.1 (was v2.56.0)
• Release notes:

# Release v2.57.1

Generating release notes from v2.57.0 to HEAD
### 📝 Other Changes

- fix(credentials): harden list input validation and strip data field (#816) (a5d56fa)
- fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816) (ca45d9d)

---

**Release Statistics:**
- 2 commits
- 1 contributor
- Contributors: czlonkowski

---

## Installation

### NPM Package

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 3 fix · 3 other
• Files touched: code=13 docs=2 ci=0 test=7 other=35
• 7 new commit(s) since last check
• Latest: 9e30942 — Merge pull request #830 from czlonkowski/fix/credential-pagination-816
• Recent commits:
• ced57ab [fix] fix(versions): scope workflow_versions per instance
• ed0447d [other] Merge commit from fork
• 2d70cce [chore] chore: update n8n to 2.23.0 and bump version to 2.57.0
• 56d3a29 [other] Merge pull request #815 from czlonkowski/update/n8n-2.23.0
• ca45d9d [fix] fix(credentials): paginate past 100 credentials in n8n_manage_credentials (#816)

claude-mem

• New release: v13.4.1 (was v13.3.0)
• Release notes:

## What's new

### 🟣 CMEM Online email opt-in during `npx claude-mem install`
An optional, interactive email opt-in now appears at the start of the installer. Press Enter to skip — it never blocks or fails the install.

- Collects an email + an optional "what are you working on / how can we help your team" note.
- POSTs to the live `https://cmem.ai/api/waitlist` endpoint (handles persistence, dedup, and the confirmation email server-side). Overridable via `CLAUDE_MEM_SIGNUP_URL`; tagged `source: npx-installer`.
- Skipped automatically when non-interactive, under CI, or with `CLAUDE_MEM_ONLINE_OPTIN=false`.
- Signup is persisted locally so returning users aren't re-prompted; a failed send is retried silently on the next install.
- No secrets ship in the npx package — the endpoint is unauthenticated and the Resend key stays server-side. The waitlist endpoint was extended to capture the optional note.

### 🔴 Fixes
- Remove a duplicate `ModeManager` import that was breaking the typecheck.
- Exempt the `transcript-watcher-entry` CLI process entry point from the console-logging guard.

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 5 docs · 5 feat · 36 fix · 7 other · 2 refactor · 1 test
• Files touched: code=117 docs=27 ci=4 test=86 other=30
• 58 new commit(s) since last check
• Latest: f267d1d — docs: regenerate CHANGELOG for v13.4.1
• Recent commits:
• c46563c [docs] docs: add plan-07..11 architectural fix plans
• 61fe70a [refactor] refactor: remove retry/observation queue, replace with SessionMessageBuffer
• 7a6eaca [fix] fix(build): plan-10 artifact hygiene — green typecheck, drop better-auth from worker, CI gate
• a8cf671 [fix] fix(tests+pipeline): repair test suite (46->7 fails) + plan-09 schema self-healing
• 90f7049 [fix] fix(opencode+openclaw): plan-08 event-contract correctness — capture actually works

voicemode

• New release: v8.6.2 (was v8.6.1)
• Release notes:

### Fixed

- **Kokoro service no longer silently dies after `UVICORN_LIMIT_MAX_REQUESTS` requests on Linux** (VM-1398, [#448](https://github.com/mbailey/voicemode/issues/448) — reported by @JulesGosnell) — The systemd unit `voicemode-kokoro.service` combined `Restart=on-failure` with `Environment="UVICORN_LIMIT_MAX_REQUESTS=25"` (the memory-leak workaround from VM-358). Hitting the request limit makes uvicorn exit cleanly (status 0), which `on-failure` treats as a successful shutdown — so the service stayed dead and the 26th TTS request failed with connection refused. The leak workaround was silently disabling the very service it was meant to protect. **Switched to `Restart=always`** in the kokoro template (and defensively in the whisper and serve templates for the same shape of bug). `RestartPreventExitStatus=127` still guards against fight-looping on a missing executable. **macOS / launchd was unaffected** — its plist uses `KeepAlive=true`, which restarts on clean exits too. To recover an existing v1.3.0 deployment without reinstalling: `sed -i 's/^Restart=on-failure/Restart=always/' ~/.config/systemd/user/voicemode-kokoro.service && systemctl --user daemon-reload && systemctl --user restart voicemode-kokoro.service`.

- **Fresh plugin installs no longer fail with "voicemode-mcp-launcher not found"** (VM-1364, [#441](https://github.com/mbailey/voicemode/issues/441)) — `.mcp.json` on `master` was switched to invoke the new `voicemode-mcp-launcher` binary (VM-1314), but the binary only exists in `master` — it isn't shipped in the latest PyPI release (8.6.1). Because Claude Code plugins install `.mcp.json` from the default branch, every fresh `claude plugin install voicemode@voicemode` produced `voicemode: voicemode-mcp-launcher - ✗ Failed to connect`. Reverted `.mcp.json` to the v8.6.1 form (`command: uv`, `args: [run, voicemode]`) so plugin installs work against the released package. The launcher invocation will be restored at the next release, alongside the published binary.
- **Provider tests no longer fail when user voicemode.env overrides TTS/STT base URLs** (VM-1138, [#370](https://github.com/mbailey/voicemode/pull/370)) -- `tests/test_providers.py` (and friends) now isolate themselves from the user's `~/.voicemode/voicemode.env`, so `VOICEMODE_TTS_BASE_URLS` / `VOICEMODE_STT_BASE_URLS` overrides on the dev machine no longer break the suite.
- **Opus playback no longer truncates the tail** -- Pad trailing silence on opus output to prevent the last frame being clipped.
- **Kokoro install pinned to fork with opus tail fix** -- `voicemode service install kokoro` now installs from `ai-cora/Kokoro-FastAPI` until the fix lands upstream.


## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 3 docs · 3 feat · 4 fix · 14 other
• Files touched: code=0 docs=4 ci=1 test=5 other=16
• 28 new commit(s) since last check
• Latest: a12b1fa — docs(changelog): add missing Unreleased entries for VM-1024 (autofocus) and VM-1421 (install.sh STT/TTS split)
• Recent commits:
• c183c91 [other] deps(deps): bump starlette from 0.47.3 to 0.49.1
• a3608c6 [other] Merge pull request #438 from mbailey/dependabot/uv/starlette-0.49.1
• 276f1be [chore] chore(dependabot): group every bump (major/minor/patch) into one weekly PR
• d0a7629 [other] Merge branch 'chore/dependabot-group-all-bumps'
• af7232b [fix] fix(VM-1398): systemd Restart=always so kokoro survives UVICORN_LIMIT_MAX_REQUESTS

playwright-mcp

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore
• Files touched: code=0 docs=0 ci=0 test=0 other=1
• 1 new commit(s) since last check
• Latest: 3201718 — chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)
• Recent commits:
• 3201718 [chore] chore(deps-dev): bump qs from 6.15.1 to 6.15.2 (#1627)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.5.1 (was ctx7@0.4.5)
• Release notes:

### Patch Changes

- ea91d7d: `ctx7 login` now always uses the device-code flow. The localhost-callback path is removed — every install (laptop, SSH, Codespace, Docker, CI) goes through the same boxed prompt and verification page. Drops the `--device` flag (it was the opt-in for what's now the default). Older CLI versions (≤ 0.5.0) continue to work against the unchanged auth endpoints, so pinned installs are unaffected.

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore · 1 ci · 1 docs · 3 feat · 5 fix · 1 other
• Files touched: code=23 docs=26 ci=7 test=6 other=1
• 14 new commit(s) since last check
• Latest: 0807914 — ci: update GitHub Actions to current major versions (#2739)
• Recent commits:
• 5a180d5 [feat] feat(cli): OAuth 2.0 device authorization flow (#2714)
• 22a6089 [chore] chore(release): version packages (#2719)
• ea91d7d [feat] feat(cli): device-code flow is now the default for ctx7 login (#2720)
• d82bcc9 [chore] chore(release): version packages (#2721)
• 0a6ecdc [other] docs pages

lightrag

• New release: v1.5.1 (was v1.5.0)
• Release notes:

## Importance Notes

To prevent weaker LLMs from inadvertently incorporating few-shot examples from the prompt into the extraction results, the official **entity-relation extraction prompt has been updated in the current version by removing the few-shot examples**. Empirical evaluation on the Qwen3-30B model shows no degradation in extraction quality. However, users are advised to monitor performance carefully on their specific LLMs to ensure consistent results.

To modify the system's default behavior, provide entity types and example prompt texts via env var `ENTITY_TYPE_PROMPT_FILE`.

## What's New

* feat(query): optionally send chunk heading path to the LLM by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3211
* ✨ feat(chunker): inject split-table headers at split time, freeze split slices by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3218
* feat(extract): inject section heading breadcrumb into extraction prompts by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3225
* Add query cancellation with Stop button by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3227
* ♻️ refactor: make extraction prompt examples format-only by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3231

## What's Changed

* build(deps): update redis requirement from <8.0.0,>=5.0.0 to >=5.0.0,<9.0.0 by @dependabot[bot] in https://github.com/HKUDS/LightRAG/pull/3194
* 🔧 chore(mineru): default MINERU_LOCAL_IMAGE_ANALYSIS to false by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3208
* build(deps): bump the react group in /lightrag_webui with 3 updates by @dependabot[bot] in https://github.com/HKUDS/LightRAG/pull/3199
* build(deps): bump lucide-react from 1.16.0 to 1.17.0 in /lightrag_webui in the ui-components group by @dependabot[bot] in https://github.com/HKUDS/LightRAG/pull/3200

• Impact: 🟡 review (code touched)
• Commit mix: 5 build · 2 chore · 1 docs · 8 feat · 13 fix · 69 other · 2 refactor · 1 test
• Files touched: code=0 docs=14 ci=2 test=23 other=40
• 101 new commit(s) since last check
• Latest: 4bbb709 — 🌐 i18n(locale): add entity labels
• Recent commits:
• 5c25331 [build] build(deps): bump the react group in /lightrag_webui with 3 updates
• 46195b6 [build] build(deps): bump lucide-react
• c05e2f5 [build] build(deps-dev): bump the build-tools group
• a62fa0a [fix] fix: honor postgres search path in table checks
• fdd6ee0 [other] ⬆️ chore(deps): bump redis to 8.0.0

everything-claude-code

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 7 docs · 17 feat · 26 fix · 2 other · 2 test
• Files touched: code=0 docs=262 ci=8 test=0 other=30
• 55 new commit(s) since last check
• Latest: edebcc8 — feat(discord): release -> #announcements auto-post + pin + GitHub Discussions (#2201)
• Recent commits:
• bc8e12b [feat] feat: add dynamic workflow team orchestration surface
• ab5e17f [feat] feat: extend session-adapter layer with codex-worktree + opencode adapters (#2145)
• 7113b5b [feat] feat: MCP inventory (ecc.mcp.v1) — unified cross-harness MCP config view (#2146)
• c8caf19 [feat] feat: worktree-lifecycle service (deterministic conflict prediction + safe GC) (#2164)
• 898fd23 [fix] fix: guard two script edge cases (tolerant package.json parse, set -u empty array) (#2088)

claude-mem

• New release: v13.4.2 (was v13.4.0)
• Release notes:

## What's new

**Installer: \"work email\" opt-in** — the CMEM Online signup prompt in `npx claude-mem install` now asks for your *work* email (placeholder `you@company.com`). This surfaces which orgs are adopting claude-mem.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 2 docs · 1 feat · 23 fix · 5 other
• Files touched: code=38 docs=20 ci=3 test=21 other=26
• 33 new commit(s) since last check
• Latest: f7f82e2 — docs: regenerate CHANGELOG for v13.4.2
• Recent commits:
• cf450ce [fix] fix(build): enforce shipped dependency-closure boundary (plan-10, closes #2783) (#2800)
• 9efb04d [fix] fix(hooks): bun-runner.js parses on pre-ES2020 Node (closes #2791) (#2801)
• 3415d97 [fix] fix(context): BMP-only injected context to stop astral-surrogate 400s (closes #2787) (#2802)
• 530b2a0 [fix] fix(pipeline): don't drop observations when the prompt row is merely absent (closes #2794) (#2803)
• cc0ce51 [fix] fix(grammar): plain-JS query + custom-grammar loading (plan-13, closes #2786) (#2805)

voicemode

• New release: v8.7.1 (was v8.6.2)
• Release notes:

## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services
uvx voice-mode-install

# Add to Claude Code MCP
claude mcp add --scope user voicemode -- uvx --refresh voice-mode

Alternative: Direct Installation

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 3 docs · 2 feat · 2 fix · 5 other · 1 test
• Files touched: code=0 docs=2 ci=0 test=5 other=8
• 14 new commit(s) since last check
• Latest: 915ee90 — Merge branch 'feat/VM-1499-make-converse-transcript-echo-reliable-dont'
• Recent commits:
• 004e725 [feat] feat(VM-1421): split STT/TTS in install.sh -- whisper.cpp for STT, choose mlx-audio or Kokoro for TTS
• 3c9d4ff [fix] fix(VM-1466): trust X-Forwarded-For only from configured proxies (GHSA-2qvv-vjq9-g5r4)
• 612dba9 [fix] fix(VM-291): don't discard long utterances ending in a trigger word
• c2a791c [other] Merge branch 'feat/VM-1421-installsh-split-stttts-whispercpp-always-for-stt'
• ef08fd5 [other] Merge branch 'fix/VM-291-prevent-data-loss-when-repeat-trigger-words'

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.

[github-actions]

Upstream Dependency Updates

The following upstream repos have new changes that may affect Maestro:

context7

• New release: ctx7@0.5.1 (was ctx7@0.4.5)
• Release notes:

### Patch Changes

- ea91d7d: `ctx7 login` now always uses the device-code flow. The localhost-callback path is removed — every install (laptop, SSH, Codespace, Docker, CI) goes through the same boxed prompt and verification page. Drops the `--device` flag (it was the opt-in for what's now the default). Older CLI versions (≤ 0.5.0) continue to work against the unchanged auth endpoints, so pinned installs are unaffected.

• Impact: 🟡 review (code touched)
• Commit mix: 4 chore · 1 ci · 1 docs · 4 feat · 5 fix · 3 other
• Files touched: code=23 docs=36 ci=9 test=6 other=6
• 18 new commit(s) since last check
• Latest: 1f6212b — trigger the Test workflow only when they touch packages/** or the root files that affect the toolchain (#2747)
• Recent commits:
• 5a180d5 [feat] feat(cli): OAuth 2.0 device authorization flow (#2714)
• 22a6089 [chore] chore(release): version packages (#2719)
• ea91d7d [feat] feat(cli): device-code flow is now the default for ctx7 login (#2720)
• d82bcc9 [chore] chore(release): version packages (#2721)
• 0a6ecdc [other] docs pages

lightrag

• New release: v1.5.1 (was v1.5.0)
• Release notes:

## Importance Notes

To prevent weaker LLMs from inadvertently incorporating few-shot examples from the prompt into the extraction results, the official **entity-relation extraction prompt has been updated in the current version by removing the few-shot examples**. Empirical evaluation on the Qwen3-30B model shows no degradation in extraction quality. However, users are advised to monitor performance carefully on their specific LLMs to ensure consistent results.

To modify the system's default behavior, provide entity types and example prompt texts via env var `ENTITY_TYPE_PROMPT_FILE`.

## What's New

* feat(query): optionally send chunk heading path to the LLM by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3211
* ✨ feat(chunker): inject split-table headers at split time, freeze split slices by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3218
* feat(extract): inject section heading breadcrumb into extraction prompts by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3225
* Add query cancellation with Stop button by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3227
* ♻️ refactor: make extraction prompt examples format-only by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3231

## What's Changed

* build(deps): update redis requirement from <8.0.0,>=5.0.0 to >=5.0.0,<9.0.0 by @dependabot[bot] in https://github.com/HKUDS/LightRAG/pull/3194
* 🔧 chore(mineru): default MINERU_LOCAL_IMAGE_ANALYSIS to false by @danielaskdd in https://github.com/HKUDS/LightRAG/pull/3208
* build(deps): bump the react group in /lightrag_webui with 3 updates by @dependabot[bot] in https://github.com/HKUDS/LightRAG/pull/3199
* build(deps): bump lucide-react from 1.16.0 to 1.17.0 in /lightrag_webui in the ui-components group by @dependabot[bot] in https://github.com/HKUDS/LightRAG/pull/3200

• Impact: 🟡 review (code touched)
• Commit mix: 5 build · 2 chore · 1 docs · 8 feat · 13 fix · 72 other · 2 refactor · 1 test
• Files touched: code=0 docs=14 ci=2 test=25 other=41
• 104 new commit(s) since last check
• Latest: fa213a8 — Bump core version to 1.5.2 and API version to 0380
• Recent commits:
• 5c25331 [build] build(deps): bump the react group in /lightrag_webui with 3 updates
• 46195b6 [build] build(deps): bump lucide-react
• c05e2f5 [build] build(deps-dev): bump the build-tools group
• a62fa0a [fix] fix: honor postgres search path in table checks
• fdd6ee0 [other] ⬆️ chore(deps): bump redis to 8.0.0

everything-claude-code

• New release: v2.0.0 (was v1.10.0)
• Release notes:

# ECC 2.0.0 — The Agent Harness Operating System

ECC 2.0.0 is the stable graduation of the 2.0 line: ECC as a cross-harness operating system for agentic work. Claude Code stays first-class; Codex, OpenCode, Cursor, Gemini, Zed, and terminal-only workflows share the same skills, rules, hooks, MCP conventions, release gates, and operator workflows. This is the months-in-the-making release the rc.1 candidate was building toward.

## The control pane (early build)

![ECC Control Pane — sessions, metrics, and the work-items board](https://github.com/affaan-m/ECC/releases/download/v2.0.0/ecc-pane-1.png)

![Operator column — knowledge recall, connectors, and runnable agent actions](https://github.com/affaan-m/ECC/releases/download/v2.0.0/ecc-pane-2.png)

The ECC 2.0 operating surface, running locally (`node scripts/control-pane.js`): operator recall search, live session metrics, a work-items board with ready/running/blocked/done lanes, and an operator column that drives knowledge sync/recall, graph backfill, and the TUI. The session adapters and MCP inventory below feed this board. Build-in-public continues in the Discord #control-pane channel.

## What is new

- **261 public skills** across coding, research, security, media, enterprise ops, and agent workflows (rc.1 shipped 243) — plus 64 agents and 84 commands.
- **The control-pane substrate** — the foundation of the ECC 2.0 operating surface:
  - harness-neutral session adapters (`ecc.session.v1`) covering Claude Code, Codex, OpenCode, and dmux — one schema for "which agent is where, doing what"
  - MCP inventory (`ecc.mcp.v1`) — one normalized view of every MCP server config across harnesses, with fragmentation/drift detection and secret redaction (it caught a real arg-carried key leak during development)
  - worktree-lifecycle service — deterministic merge-conflict prediction and safe GC for parallel agent worktrees
- **`orch-*` orchestrator skill family** plus dynamic workflow team orchestration — multi-agent fan-out as a first-class surface.

• Impact: 🟡 review (code touched)
• Commit mix: 1 chore · 3 ci · 10 docs · 18 feat · 26 fix · 4 other · 2 test
• Files touched: code=0 docs=248 ci=11 test=0 other=43
• 64 new commit(s) since last check
• Latest: c888d2b — docs: update Greptile sponsor placement
• Recent commits:
• bc8e12b [feat] feat: add dynamic workflow team orchestration surface
• ab5e17f [feat] feat: extend session-adapter layer with codex-worktree + opencode adapters (#2145)
• 7113b5b [feat] feat: MCP inventory (ecc.mcp.v1) — unified cross-harness MCP config view (#2146)
• c8caf19 [feat] feat: worktree-lifecycle service (deterministic conflict prediction + safe GC) (#2164)
• 898fd23 [fix] fix: guard two script edge cases (tolerant package.json parse, set -u empty array) (#2088)

n8n-mcp

• New release: v2.57.2 (was v2.57.1)
• Release notes:

# Release v2.57.2

Generating release notes from v2.57.1 to HEAD
### 📝 Other Changes

- fix(workflow): don't let a null settings payload wipe current settings (75887af)
- fix(workflow): object-merge settings on full update so partial payloads don't drop keys (0a07f2f)
- test(workflow): cover nodeGroups echo-back in allowlist (supersedes #835, #839) (33e3d5b)
- fix(workflow): allowlist write payload so n8n 2.x echo-back fields don't break updates (109c483)

---

**Release Statistics:**
- 4 commits
- 1 contributor
- Contributors: czlonkowski

---

## Installation

• Impact: 🟡 review (code touched)
• Commit mix: 7 fix · 2 other · 1 test
• Files touched: code=4 docs=1 ci=0 test=4 other=13
• 10 new commit(s) since last check
• Latest: 8baa226 — Merge pull request #836 from cnYui/codex/normalize-http-mcp-workflow-input
• Recent commits:
• 109c483 [fix] fix(workflow): allowlist write payload so n8n 2.x echo-back fields don't break updates
• 33e3d5b [test] test(workflow): cover nodeGroups echo-back in allowlist (supersedes #835, #839)
• 0a07f2f [fix] fix(workflow): object-merge settings on full update so partial payloads don't drop keys
• 75887af [fix] fix(workflow): don't let a null settings payload wipe current settings
• 7fec791 [other] Merge pull request #840 from czlonkowski/fix/update-workflow-additional-properties

claude-mem

• New release: v13.5.5 (was v13.4.0)
• Release notes:

## Telemetry Reliability Signals (Plan 14)

claude-mem instrumented success well — failure was invisible. This release adds the five highest-value missing reliability signals (#2874). Everything is closed-enum/count-only, whitelisted in the scrubber, and disclosed in both the [public docs](https://docs.claude-mem.ai/telemetry) and `claude-mem telemetry`.

### Search retrieval quality (`search_performed`)
- `result_count`, `search_strategy` (`chroma|fts|filter_only`), `chroma_available`, `fallback_reason` (`none|chroma_connection|chroma_error|chroma_not_initialized`)
- Zero-result rate is now computable, and Chroma's silent degradation to FTS is visible.

### Compression trust (`session_compressed`)
- `fabrication_detected` / `fabricated_count` — commit-hash fabrication by the observer model, on every emit path
- Respawn-gated invalid-output events: `invalid_output_class` (`xml|idle|prose|poisoned`), `consecutive_invalid_outputs`, `respawn_triggered`
- `outcome: aborted` with `abort_reason` (`idle|shutdown|overflow|restart_guard|quota|poisoned|none`), emitted where all abort flows converge

### Worker lifecycle
- New `worker_stopped` event: `uptime_seconds`, `shutdown_reason` (`stop|restart|signal`)
- Crash detection via clean-shutdown sentinel: `worker_started` now reports `previous_shutdown` (`crash|clean|unknown`) and `previous_uptime_seconds`
- Memory health: integer `process_rss_mb` / `heap_used_mb` on lifecycle events and the heartbeat

### Hook failures
- New `hook_failed` event over the direct CLI transport (the worker being unreachable IS the failure being reported), threshold-gated on the fail-loud counter and awaited before process exit so events survive short-lived hook processes

• Impact: 🟡 review (code touched)
• Commit mix: 8 chore · 9 docs · 12 feat · 29 fix · 11 other · 1 refactor · 1 test
• Files touched: code=71 docs=61 ci=3 test=41 other=30
• 71 new commit(s) since last check
• Latest: ec86333 — docs: regenerate changelog for v13.5.5
• Recent commits:
• cf450ce [fix] fix(build): enforce shipped dependency-closure boundary (plan-10, closes #2783) (#2800)
• 9efb04d [fix] fix(hooks): bun-runner.js parses on pre-ES2020 Node (closes #2791) (#2801)
• 3415d97 [fix] fix(context): BMP-only injected context to stop astral-surrogate 400s (closes #2787) (#2802)
• 530b2a0 [fix] fix(pipeline): don't drop observations when the prompt row is merely absent (closes #2794) (#2803)
• cc0ce51 [fix] fix(grammar): plain-JS query + custom-grammar loading (plan-13, closes #2786) (#2805)

voicemode

• New release: v8.7.1 (was v8.6.2)
• Release notes:

## Installation

### Quick Start (Recommended)

```bash
# Install UV package manager (if not already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install VoiceMode and configure services
uvx voice-mode-install

# Add to Claude Code MCP
claude mcp add --scope user voicemode -- uvx --refresh voice-mode

Alternative: Direct Installation

• Impact: 🟡 review (code touched)
• Commit mix: 2 chore · 3 docs · 2 feat · 4 fix · 7 other · 1 test
• Files touched: code=0 docs=2 ci=0 test=7 other=9
• 19 new commit(s) since last check
• Latest: 9616595 — Merge branch 'fix/VM-1556-default-generated-voicemodeenv-silently-routes-to'
• Recent commits:
• 004e725 [feat] feat(VM-1421): split STT/TTS in install.sh -- whisper.cpp for STT, choose mlx-audio or Kokoro for TTS
• 3c9d4ff [fix] fix(VM-1466): trust X-Forwarded-For only from configured proxies (GHSA-2qvv-vjq9-g5r4)
• 612dba9 [fix] fix(VM-291): don't discard long utterances ending in a trigger word
• c2a791c [other] Merge branch 'feat/VM-1421-installsh-split-stttts-whispercpp-always-for-stt'
• ef08fd5 [other] Merge branch 'fix/VM-291-prevent-data-loss-when-repeat-trigger-words'

playwright-mcp

• New release: v0.0.76 (was v0.0.75)
• Release notes:

## What's New

### New Tools

- **`browser_video_show_actions` / `browser_video_hide_actions`** — Overlay action annotations on the recorded video, or hide them again ([#40914](https://github.com/microsoft/playwright/pull/40914))

### Tool Improvements

- **Remote endpoint** — `remoteEndpoint` now accepts a `ConnectOptions` object, not just a URL string ([#40964](https://github.com/microsoft/playwright/pull/40964))
- **`--output-max-size`** — Cap the size of tool responses, with post-response disk eviction of oversized output ([#41031](https://github.com/microsoft/playwright/pull/41031))
- **`--browser`** — Support `moz-firefox` BiDi channels ([#41126](https://github.com/microsoft/playwright/pull/41126))

## Bug Fixes

- Support `remoteHeaders` for the remote browser endpoint ([#40828](https://github.com/microsoft/playwright/pull/40828), [#41156](https://github.com/microsoft/playwright/pull/41156))
- Use `waitUntil: 'commit'` when navigating back/forward ([#41153](https://github.com/microsoft/playwright/pull/41153))
- Report invalid tool arguments instead of failing opaquely ([#40979](https://github.com/microsoft/playwright/pull/40979))
- Use a writable cache directory for MCP user data instead of the browsers path ([#40961](https://github.com/microsoft/playwright/pull/40961))
- Disconnect the tracked browser when the browser tracker is disposed ([#40967](https://github.com/microsoft/playwright/pull/40967))
- Report a missing `ffmpeg` distinctly from a missing browser ([#40867](https://github.com/microsoft/playwright/pull/40867))

• Impact: 🟡 review (code touched)
• Commit mix: 3 chore
• Files touched: code=0 docs=1 ci=0 test=0 other=4
• 3 new commit(s) since last check
• Latest: b301c37 — chore: mark v0.0.76 (#1649)
• Recent commits:
• aa55fbb [chore] chore(deps-dev): bump hono from 4.12.18 to 4.12.23 (#1638)
• 40de619 [chore] chore: roll Playwright to 1.61.0-alpha-1781023400000 (#1648)
• b301c37 [chore] chore: mark v0.0.76 (#1649)

---

Action Required

Review these changes and determine if Maestro needs updating:

• Check if new superpowers skills should be orchestrated
• Check if Context7 API has changed
• Check if new Vercel agent skills are available
• Check if Playwright MCP API or capabilities have changed
• Check if claude-mem MCP tool names or hook schema have changed
• Check if UI UX Pro Max, n8n-MCP, VoiceMode, Everything Claude Code, LightRAG, Karpathy Skills, or Caveman have breaking changes
• Update SKILL.md if needed
• Update checklists if needed
• Bump version in plugin.json if changes made

This issue was automatically created by the upstream tracker workflow.