feat(foundry): add creator agent and create skill; rename self-mentor → curator

- rename foundry:self-mentor agent to foundry:curator across all files (agent file, audit template, codex mirror, 23+ cross-ref files)
- add foundry:creator agent for scaffolding new agent definitions via /create skill
- add foundry:create skill with full workflow for new-agent scaffolding
- add brainstorm tree-viewer.html live visualization script

---
Co-authored-by: Claude Code <noreply@anthropic.com>

Author: Borda

.claude/README.md

@@ -89,7 +89,7 @@ These entries pre-authorize `Read`, `Glob`, `Grep`, and `Write` on directories t
 | Permission              | Description                                   | Typical use case                                                                                                                                                                                               |
 | ----------------------- | --------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
 | `Read(.claude/*.md)`    | Read top-level `.claude/` markdown files      | Agents read CLAUDE.md, permissions-guide.md, and TEAM_PROTOCOL.md at spawn                                                                                                                                     |
-| `Read(.claude/**/*.md)` | Read any nested `.claude/` markdown file      | Agents and skills read their own agent/skill/rule files; self-mentor reads config files for audit                                                                                                              |
+| `Read(.claude/**/*.md)` | Read any nested `.claude/` markdown file      | Agents and skills read their own agent/skill/rule files; curator reads config files for audit                                                                                                                  |
 | `Read(.claude/logs/**)` | Read log files under `.claude/logs/`          | `/calibrate` reads calibrations.jsonl for historical context; `/audit` reads audit-errors.jsonl                                                                                                                |
 | `Edit(.claude/logs/**)` | Edit log files under `.claude/logs/`          | Skills append to calibrations.jsonl and audit-errors.jsonl without Bash redirection                                                                                                                            |
 | `Read(./**)`            | Read any file in the project root             | Teammates read `TEAM_PROTOCOL.md` and agent files at spawn; skills read their own SKILL.md files                                                                                                               |
@@ -143,7 +143,7 @@ ______________________________________________________________________
 | `Bash(mkdir -p .plans/closed/)`       | Create `.plans/closed/` directory                   | Plans are moved here when completed                                                                       |
 | `Bash(mkdir -p .reports/calibrate/*)` | Create `.reports/calibrate/` skill run subdirs      | `/calibrate` creates a timestamped run dir before spawning pipeline agents                                |
 | `Bash(mkdir -p .reports/resolve/*)`   | Create `.reports/resolve/` skill run subdirs        | `/resolve` creates a run dir for lint+QA gate artifacts                                                   |
-| `Bash(mkdir -p .reports/audit/*)`     | Create `.reports/audit/` skill run subdirs          | `/audit` creates a timestamped run dir before spawning self-mentor agents                                 |
+| `Bash(mkdir -p .reports/audit/*)`     | Create `.reports/audit/` skill run subdirs          | `/audit` creates a timestamped run dir before spawning curator agents                                     |
 | `Bash(mkdir -p .reports/review/*)`    | Create `.reports/review/` skill run subdirs         | `/oss:review` creates a run dir for multi-agent review artifacts                                          |
 | `Bash(mkdir -p .reports/analyse/*)`   | Create `.reports/analyse/` skill run subdirs        | `/analyse` creates subdirs for thread, ecosystem, and health modes                                        |
 | `Bash(mkdir -p .experiments/*)`       | Create `.experiments/` skill run subdirs            | `/optimize` creates a run dir for run mode artifacts                                                      |

.claude/permissions-guide.md

@@ -137,14 +137,14 @@ Parent agent responsibilities:
 - `oss-shepherd`: issue triage, maintainer review, SemVer and release governance
 - `solution-architect`: architecture planning, API contracts, migration design
 - `web-explorer`: authoritative external docs/changelogs/API delta research
-- `self-mentor`: configuration drift, instruction overlap, calibration/gate hygiene
+- `curator`: configuration drift, instruction overlap, calibration/gate hygiene
 
 ### Collaboration team patterns
 
 - Architecture/public API changes: `solution-architect` + `sw-engineer` + `qa-specialist` + `doc-scribe`
 - Security-sensitive features: `security-auditor` + `sw-engineer` + `qa-specialist`
 - Data pipeline changes: `data-steward` + `sw-engineer` + `qa-specialist`
-- Toolchain/CI quality changes: `ci-guardian` + `linting-expert` + `self-mentor`
+- Toolchain/CI quality changes: `ci-guardian` + `linting-expert` + `curator`
 - External migration/release-note driven changes: `web-explorer` + `solution-architect` + `sw-engineer`
 - Release readiness: `oss-shepherd` + `ci-guardian` + `doc-scribe` + `qa-specialist`
 
@@ -237,7 +237,7 @@ Parent agent responsibilities:
 - Exact references and source-backed evidence are required for decisions
 - Volatile ecosystem/tooling behavior could invalidate cached assumptions
 
-### Spawn `self-mentor` when:
+### Spawn `curator` when:
 
 - Config/skill/agent drift or duplication is suspected
 - Routing quality, calibration leakage, or weak gate coverage is reported

.codex/AGENTS.md

@@ -66,7 +66,7 @@ All agents are standardized on `gpt-5.4-mini`. Differentiation is via reasoning
 | **oss-shepherd**       | high   | Issue triage, PR review, SemVer, pyDeprecate, release checklist         |
 | **solution-architect** | high   | System design, ADRs, API compatibility, migration planning              |
 | **web-explorer**       | medium | External docs/release-note extraction and evidence gathering            |
-| **self-mentor**        | medium | Config quality checks, drift/leak detection, workflow hygiene           |
+| **curator**            | medium | Config quality checks, drift/leak detection, workflow hygiene           |
 
 ### Spawn rules
 
@@ -102,7 +102,7 @@ Reasoning allocation:
 | ---------- | --------------------------------------------------------------------- | ------------------------------------------------------ |
 | **xhigh**  | qa-specialist, security-auditor                                       | Adversarial: exhaustive search for what could go wrong |
 | **high**   | sw-engineer, squeezer, data-steward, oss-shepherd, solution-architect | Analytical: depth without unbounded budget             |
-| **medium** | doc-scribe, ci-guardian, linting-expert, web-explorer, self-mentor    | Writing/config/research balance                        |
+| **medium** | doc-scribe, ci-guardian, linting-expert, web-explorer, curator        | Writing/config/research balance                        |
 
 ### Profiles
 
@@ -153,7 +153,7 @@ Agent targeting examples:
 ```text
 use the qa-specialist to review tests/ for missing edge cases
 use the solution-architect to produce a minimal migration plan for this API change
-use the self-mentor to review .codex drift and weak gates
+use the curator to review .codex drift and weak gates
 ```
 
 ## 🪙 RTK Integration

.codex/README.md

@@ -1,12 +1,12 @@
-name                    = "self-mentor"
+name                    = "curator"
 model                   = "gpt-5.4-mini"
 model_reasoning_effort  = "xhigh"
 model_reasoning_summary = "detailed"
 approval_policy         = "on-request"
 sandbox_mode            = "workspace-write"
 
 developer_instructions = """
-You are the Self Mentor subagent for Borda Codex config quality.
+You are the Curator subagent for Borda Codex config quality.
 Philosophy: "Keep instructions lean, current, and consistent."
 
 ## Scope

.codex/agents/self-mentor.toml .codex/agents/curator.toml.codex/agents/self-mentor.toml renamed to .codex/agents/curator.toml

@@ -1,6 +1,6 @@
 {
   "agents": {
-    "self-mentor": [
+    "curator": [
       "Review Checklist",
       "calibration coverage"
     ],

.codex/calibration/benchmarks.json

@@ -11,7 +11,7 @@ HOME_CFG="$HOME/.codex/config.toml"
 TASKS="$ROOT/.codex/calibration/tasks.json"
 BENCHMARKS="$ROOT/.codex/calibration/benchmarks.json"
 SKILLS=(review develop resolve audit calibrate release investigate sync manage analyse optimize research)
-AGENTS=(sw-engineer qa-specialist squeezer doc-scribe security-auditor data-steward ci-guardian linting-expert oss-shepherd solution-architect web-explorer self-mentor)
+AGENTS=(sw-engineer qa-specialist squeezer doc-scribe security-auditor data-steward ci-guardian linting-expert oss-shepherd solution-architect web-explorer curator)
 
 LEAKS=0
 FAILS=0

.codex/calibration/run.sh

@@ -148,6 +148,6 @@ config_file = "agents/solution-architect.toml"
 description = "Web and external-doc research: package versions, changelogs, migration guides, release notes, API reference extraction."
 config_file = "agents/web-explorer.toml"
 
-[agents.self-mentor]
+[agents.curator]
 description = "Configuration quality reviewer: detect duplication, drift, stale references, weak gates, and instruction hygiene issues."
-config_file = "agents/self-mentor.toml"
+config_file = "agents/curator.toml"