feat(launch): AI-managed launch automation toolkit

.github/issue-templates/good-first-issue-1.yml

@@ -0,0 +1,18 @@
+title: "Add a `glyph_render --watch` mode that re-renders on spec change"
+labels: [good first issue, cli]
+body: |
+  ## Goal
+  Add a `--watch` flag to the `glyph render` CLI so a JSON spec edit triggers an
+  automatic re-render (with the existing byte-stable pipeline).
+
+  ## Why
+  Tightens the agent ↔ human authoring loop. Pairs with the playground.
+
+  ## Hints
+  - Entry point: `packages/cli/src/render.ts`
+  - Use `chokidar` (already a transitive dep) or `node:fs.watch`.
+  - Re-run the same `compileSpec` + `renderSvg` path; print the new SHA-256.
+
+  ## Definition of done
+  - `glyph render spec.json --watch` re-renders on save.
+  - Test under `packages/cli/__tests__/watch.test.ts`.

.github/issue-templates/good-first-issue-2.yml

@@ -0,0 +1,20 @@
+title: "New audit rule: detect overlapping axis labels"
+labels: [good first issue, auditor]
+body: |
+  ## Goal
+  Add an audit rule that flags overlapping x-axis or y-axis tick labels at the
+  rendered SVG layer.
+
+  ## Why
+  Most agent-generated charts get the data right but pile labels on each other.
+  Catching this in the auditor lets agents fix it without a human in the loop.
+
+  ## Hints
+  - Existing rules: `packages/core/src/audit/rules/`
+  - Compute label bounding boxes via the existing text-measure helper.
+  - Threshold: any horizontal overlap > 0 px on the major-axis ticks.
+
+  ## Definition of done
+  - New rule file `axis-label-overlap.ts` with unit tests.
+  - Listed in `docs/AUDIT.md`.
+  - Bumped rule count in README ("11 → 12").

.github/issue-templates/good-first-issue-3.yml

@@ -0,0 +1,18 @@
+title: "New brand preset: `theme: \"oss-newsletter\"`"
+labels: [good first issue, theming]
+body: |
+  ## Goal
+  Add a fifth brand preset suited to OSS newsletters and dev.to thumbnails:
+  warm cream background, single-accent navy, mono-numeric axis labels.
+
+  ## Why
+  Lowers the activation energy for community writeups.
+
+  ## Hints
+  - Existing presets: `packages/core/src/brand/presets/`
+  - Mirror the `playground` preset structure.
+
+  ## Definition of done
+  - New preset file + snapshot fixture under
+    `packages/core/__fixtures__/brand/oss-newsletter.svg`.
+  - Documented in `site/index.html#themes`.

.github/issue-templates/good-first-issue-4.yml

@@ -0,0 +1,18 @@
+title: "Add Vega-Lite → Glyph translator quickstart in docs/LEARN.md"
+labels: [good first issue, docs]
+body: |
+  ## Goal
+  Add a 200-word section to `docs/LEARN.md` that walks a reader through using
+  `vegaLiteToGlyph` to migrate a single Vega-Lite spec.
+
+  ## Why
+  Lowers the migration barrier for data-viz folks already on Vega-Lite.
+
+  ## Hints
+  - Translator lives in `packages/core/src/translate/vega-lite.ts`.
+  - Pick a small but interesting Vega-Lite example (e.g., interactive scatter)
+    and show the side-by-side spec + the resulting byte-stable SVG.
+
+  ## Definition of done
+  - Section added near the end of `docs/LEARN.md`.
+  - Spec + rendered SVG checked into `packages/core/__fixtures__/translate/`.

.github/issue-templates/good-first-issue-5.yml

@@ -0,0 +1,20 @@
+title: "Add LangChain `GlyphTool` to packages/integrations/langchain"
+labels: [good first issue, integrations]
+body: |
+  ## Goal
+  Ship a one-file LangChain tool wrapper around `@glyph/mcp`, exposing
+  `glyph_render` and `glyph_audit_spec` as standard tools.
+
+  ## Why
+  Makes Glyph one-line-usable from any LangChain agent.
+
+  ## Hints
+  - Use `@langchain/core/tools` `tool` helper.
+  - Spawn `npx -y @glyph/mcp` as a subprocess; talk to it over stdio.
+  - Stub a small test that renders the existing `bar.json` fixture and asserts
+    the SHA-256 hash matches.
+
+  ## Definition of done
+  - `packages/integrations/langchain/glyph_tool.py` (or `.ts`).
+  - Test that runs against the local MCP server.
+  - PR opened against the LangChain examples repo.

.github/workflows/cowork-triage.yml

@@ -0,0 +1,32 @@
+name: Cowork triage
+
+on:
+  issues:
+    types: [opened, reopened]
+  pull_request:
+    types: [opened, reopened, ready_for_review]
+
+permissions:
+  issues: write
+  pull-requests: write
+  contents: read
+
+jobs:
+  triage:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Acknowledge
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const isPR = !!context.payload.pull_request;
+            const num  = (context.payload.issue || context.payload.pull_request).number;
+            const body = isPR
+              ? "Thanks for the PR! Cowork (AI maintainer) is reviewing now. Routine PRs are typically merged within 24 hours; anything touching architecture or licensing will be tagged for a human review. See CONTRIBUTING.md → 'AI-maintained'."
+              : "Thanks for the issue! Cowork (AI maintainer) will triage within 24 hours. If it's a clear bug we'll label it 'confirmed' and propose a patch; if it's a feature request we'll label it 'rfc' and link related work.";
+            await github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo:  context.repo.repo,
+              issue_number: num,
+              body,
+            });

.gitignore

@@ -42,3 +42,8 @@ sandbox/
 # entries above only covered specific paths — this catches subdirs like
 # bindings/python/tests/__pycache__/ that pytest creates per test file).
 **/__pycache__/
+
+# Launch toolkit secrets and runtime state
+scripts/launch/.env.launch
+scripts/launch/.influencer-*.json
+scripts/launch/chrome_queue.json

CONTRIBUTING.md

@@ -13,6 +13,50 @@ If you're here to **send code**, read on.
 
 ---
 
+## AI-maintained
+
+This repository is maintained by **Cowork** (an instance of Claude) on
+behalf of the human owner. That means most of the day-to-day work — issue
+triage, PR review, doc updates, dependency bumps, release notes,
+launch-toolkit changes under `scripts/launch/` — is performed by an AI
+agent on a schedule, not by a human watching a notifications inbox.
+
+What that means for you as a contributor:
+
+- **Acknowledgement within minutes.** When you open an issue or PR, the
+  [`cowork-triage`](./.github/workflows/cowork-triage.yml) workflow posts
+  an acknowledgement immediately. That comment is from Cowork, not a
+  human, and explains what happens next.
+- **Triage within 24 hours.** Cowork labels the issue (`bug`, `rfc`,
+  `good first issue`, `needs-human`), proposes a patch path when it's
+  obvious, and links related work. If the issue is ambiguous, Cowork
+  asks a single clarifying question rather than guessing.
+- **Routine PRs reviewed automatically.** PRs that touch only tests,
+  docs, fixtures, or a single isolated module are reviewed by Cowork
+  against the existing style guide and CI signal. Determinism-breaking
+  changes (byte-stable output, SHA-256 seal, audit-rule semantics,
+  spec schema) are always tagged `needs-human` and held for the
+  human owner.
+- **Humans tagged for the things that need them.** Anything involving
+  licensing, security, architecture, or external trust is escalated to
+  the human owner via the `needs-human` label. Cowork will not merge
+  PRs with that label.
+- **Everything is auditable.** Cowork's comments are signed with
+  `— Cowork (AI maintainer)`. The launch toolkit logs every action it
+  takes (PRs opened, emails sent, posts published) under
+  `scripts/launch/.influencer-emails.json` and similar files so the
+  human owner can audit at any time.
+- **You can always ask for a human.** Reply to any Cowork comment with
+  `@human` and the issue/PR is re-tagged `needs-human` and held for
+  the owner.
+
+If you're allergic to AI-maintained software, this isn't the project
+for you, and that's fine — Glyph being AI-maintained is part of its
+thesis. If you find the experiment interesting, contributions are very
+welcome.
+
+---
+
 ## Tl;dr
 
 ```bash

scripts/launch/.env.launch.example

@@ -0,0 +1,58 @@
+# Glyph launch toolkit — copy to .env.launch and fill in.
+# This file MUST NOT be committed. Add .env.launch to your repo's .gitignore.
+
+# ---- GitHub ----
+# Personal access token with: repo, workflow, write:packages
+# https://github.com/settings/tokens
+GITHUB_TOKEN=
+
+# Repo slug — change if you fork
+REPO_SLUG=seanhanca/glyph
+
+# ---- Blogging platforms ----
+# dev.to: https://dev.to/settings/extensions
+DEVTO_API_KEY=
+
+# Hashnode: https://hashnode.com/settings/developer
+HASHNODE_TOKEN=
+HASHNODE_PUBLICATION_ID=
+
+# Medium: https://medium.com/me/settings (Integration tokens)
+MEDIUM_TOKEN=
+MEDIUM_USER_ID=
+
+# ---- Discord webhooks ----
+# Right-click channel → Integrations → Webhooks → Copy URL
+DISCORD_WEBHOOK_ANTHROPIC=
+DISCORD_WEBHOOK_MCP=
+DISCORD_WEBHOOK_LANGCHAIN=
+DISCORD_WEBHOOK_CREWAI=
+DISCORD_WEBHOOK_DUCKDB=
+DISCORD_WEBHOOK_AIENGINEER=
+DISCORD_WEBHOOK_LLAMAINDEX=
+
+# ---- Bluesky / Mastodon ----
+BLUESKY_HANDLE=
+BLUESKY_APP_PASSWORD=
+MASTODON_INSTANCE=https://fosstodon.org
+MASTODON_TOKEN=
+
+# ---- Product Hunt ----
+PRODUCTHUNT_TOKEN=
+
+# ---- Resend (newsletter pitches + cold emails) ----
+RESEND_API_KEY=
+RESEND_FROM=launch@yourdomain.tld
+
+# ---- LLM for the influencer engine (drafting replies) ----
+ANTHROPIC_API_KEY=
+# or
+OPENAI_API_KEY=
+
+# ---- YouTube (for shorts upload) ----
+YOUTUBE_OAUTH_CLIENT_JSON=./scripts/launch/.youtube-oauth.json
+
+# ---- Behavior knobs ----
+DRY_RUN=true                  # default true; flip to false when you're ready to actually post
+COLD_EMAIL_DAILY_CAP=3
+INFLUENCER_POLL_INTERVAL_HOURS=24