Menu
Use when reviewing an external repo or article to adopt practices into TAOM — security-vet first, map novel vs duplicative, port (never install), review, commit your changes.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Root-cause native Bannerlord CTDs (AccessViolation in TaleWorlds.Native.dll) via Event Log offsets, offline disassembly, and live debugger forensics. No symbols needed.
Dispatch an independent Codex verification job directly via `codex exec` Bash call, then retrieve and present results
Inspect all changed files, group by logical concern, and guide atomic per-concern commits following TAOM's 50/72 rule and commit trailer convention.
Load a snapshot saved by /context-save and present it so this session can pick up where the last one left off without re-deriving decisions.
Launch parallel deep-dive agents to review completed work for quality, standards, compatibility, completeness, and cross-system data flow
Respond to a Bannerlord engine update (Steam force-bump or deliberate migration) — preserve the decompile baseline, regen, diff, re-verify bindings and creature data, control-battle.
| name | adopt-external |
| description | Use when reviewing an external repo or article to adopt practices into TAOM — security-vet first, map novel vs duplicative, port (never install), review, commit your changes. |
| allowed-tools | Read, Write, Edit, Bash, WebFetch, WebSearch |
Evaluate an external repo or article and fold the genuinely-useful parts into TAOM — without importing bloat or risk. Thin entry point; the full procedure, principles, and accumulated gotchas live in docs/ai-includes/external-repo-adoption.md — read it first.
User shares a repo/article "to see what we can adopt." One source at a time.
Nine phases — read the full procedure in the doc before starting: Identify → Security-pass-FIRST → Map novel-vs-duplicative → Tiered recommendation → Implement (port, never install) → Adversarial review → Fix + RCA → Commit MINE only → Push.
See external-repo-adoption.md § The cycle for each phase in detail. (This skill stays thin and points to the doc rather than restating it — per external-skill-ports.md "Don't amplify a rule — point to it".)
evidence-over-claims.md A.4).python tools/audit_claude_config.py --root <repo> --external fires TAOM's SkillSpector-derived categories at full severity. Heavyweight option: NVIDIA SkillSpector run static-only in an isolated venv/Docker (never installed). It supplements, never replaces, the manual read./security-scan on our own result./security-scan) — describe the category or use audit-allow:./security-scan — runs at the end of the implement phase./deep-review, /review-codex — for the adversarial-review phase on C# changes.external-skill-ports.md, evidence-over-claims.md, simplicity-criterion.md — the rules this cycle leans on.