name	tbm-audit
description	Repo-managed Codex audit skill for C:\Dev\tbm-apps-script. Use when LT says `audit N`, `re-audit N`, `audit this PR`, `audit current state`, `attack audit`, `fact-check this PR`, `claim audit`, or `proof audit`. Preserves TBM's plain-English audit contract while delegating the evidence-bound methodology to the canonical repo skill at `C:\Dev\tbm-apps-script\.claude\skills\codex-audit\SKILL.md`.

TBM Audit

This is the repo-managed Codex wrapper for TBM audit-lane work.

It intentionally owns the installed tbm-audit skill name so the repo, not a drifting global copy, defines how TBM audits run in Codex.

Read C:\Dev\tbm-apps-script\.claude\skills\codex-audit\SKILL.md for the canonical evidence-bound methodology.
If the audit touches gate/test/review-pipeline work, also read:

C:\Dev\tbm-apps-script\ops\operating-memos\2026-04-17-agent-roles-and-audit-scope.md
C:\Dev\tbm-apps-script\ops\operating-memos\2026-04-21-builder-pre-audit-and-clean-slate-rereview.md

audit N: inspect PR N only.
re-audit N: inspect the current head of PR N as a clean-slate review.
audit current state: inspect the current checkout only.
attack audit N: switch to claim-led fact-check mode for PR N.
fact-check this PR, claim audit, proof audit: run claim-led fact-check mode against the named target or the current checkout if that is the only explicit target.
Only include stacked or dependency PR context when LT explicitly says stacked, after PR M, or equivalent.
Stay in audit lane unless LT explicitly asks for fixes.

Before relying on this wrapper, verify these files exist:

If any are missing, treat the wrapper as stale and say so explicitly instead of guessing.

The repo's .claude/skills/codex-audit/SKILL.md is the canonical audit methodology. Do not fork or restate that methodology here.
Use this wrapper to preserve Codex-specific discovery and the TBM plain-English audit command contract.
Lead with findings. Findings should prioritize bugs, regressions, weak gate logic, contract drift, and missing verification.
In Codex desktop reviews, emit ::code-comment{...} for anchored inline findings.
Do not claim tests passed unless you ran them or found current evidence that they passed.

This wrapper is for TBM audit-lane work in this repo.
It is not the skill to use for direct code changes, deploy operations, or generic prompt writing.

name	tbm-audit
description	Repo-managed Codex audit skill for C:\Dev\tbm-apps-script. Use when LT says `audit N`, `re-audit N`, `audit this PR`, `audit current state`, `attack audit`, `fact-check this PR`, `claim audit`, or `proof audit`. Preserves TBM's plain-English audit contract while delegating the evidence-bound methodology to the canonical repo skill at `C:\Dev\tbm-apps-script\.claude\skills\codex-audit\SKILL.md`.

This is the repo-managed Codex wrapper for TBM audit-lane work.

It intentionally owns the installed tbm-audit skill name so the repo, not a drifting global copy, defines how TBM audits run in Codex.

Read C:\Dev\tbm-apps-script\.claude\skills\codex-audit\SKILL.md for the canonical evidence-bound methodology.
If the audit touches gate/test/review-pipeline work, also read:

C:\Dev\tbm-apps-script\ops\operating-memos\2026-04-17-agent-roles-and-audit-scope.md
C:\Dev\tbm-apps-script\ops\operating-memos\2026-04-21-builder-pre-audit-and-clean-slate-rereview.md

audit N: inspect PR N only.
re-audit N: inspect the current head of PR N as a clean-slate review.
audit current state: inspect the current checkout only.
attack audit N: switch to claim-led fact-check mode for PR N.
fact-check this PR, claim audit, proof audit: run claim-led fact-check mode against the named target or the current checkout if that is the only explicit target.
Only include stacked or dependency PR context when LT explicitly says stacked, after PR M, or equivalent.
Stay in audit lane unless LT explicitly asks for fixes.

Before relying on this wrapper, verify these files exist:

If any are missing, treat the wrapper as stale and say so explicitly instead of guessing.

The repo's .claude/skills/codex-audit/SKILL.md is the canonical audit methodology. Do not fork or restate that methodology here.
Use this wrapper to preserve Codex-specific discovery and the TBM plain-English audit command contract.
Lead with findings. Findings should prioritize bugs, regressions, weak gate logic, contract drift, and missing verification.
In Codex desktop reviews, emit ::code-comment{...} for anchored inline findings.
Do not claim tests passed unless you ran them or found current evidence that they passed.

This wrapper is for TBM audit-lane work in this repo.
It is not the skill to use for direct code changes, deploy operations, or generic prompt writing.