Use when adding new error messages to React, or seeing "unknown error code" warnings.
repo-audit
0
0
# Install this skill:
npx skills add kotsmiltos/mk-cc-resources --skill "repo-audit"
Install specific skill from multi-skill repository
# Description
Repo audit and cross-cutting amendment protocol. Two modes β AUDIT (read-only analysis, generates codebase snapshot + pattern index + per-file reports) and AMEND (change workflow requiring snapshot and pattern lookup consultation, producing amendment records). Enforced mechanically via pre-commit hooks and CI. Use when onboarding a new repo or when making any code change that must be tracked for cross-cutting impact.
# SKILL.md
name: repo-audit
description: Repo audit and cross-cutting amendment protocol. Two modes β AUDIT (read-only analysis, generates codebase snapshot + pattern index + per-file reports) and AMEND (change workflow requiring snapshot and pattern lookup consultation, producing amendment records). Enforced mechanically via pre-commit hooks and CI. Use when onboarding a new repo or when making any code change that must be tracked for cross-cutting impact.
Provides two-mode repo governance: AUDIT generates a codebase snapshot and pattern index from scratch; AMEND enforces cross-cutting awareness for every code change via mechanical pre-commit and CI validation. Every commit touching code-like files must include a valid amendment record.
To audit a repo: say "audit" β routes to workflows/audit.md.
To make a tracked code change: say "amend" β routes to workflows/amend.md.
If unclear which mode, ask the user.
1. AUDIT mode MUST NOT modify source code. It inspects, documents, and indexes only. Outputs are documentation artifacts.
2. Any commit touching code-like files MUST include a valid amendment record. Enforced by pre-commit hooks and CI.
3. AMEND mode MUST read CLAUDE.md (codebase snapshot) and _code_audit/patterns.md (pattern index) before proposing changes.
4. Before changing code, identify the primary target, find all related implementations via the pattern index, and document what you found.
5. If the cross-cutting search finds related files you chose not to update, state why in the amendment's not_updated_files section.
6. Every amendment MUST declare integrity_check_done: true.
7. Outputs are deterministic and re-runnable. Running audit mode twice produces the same structure. Amendment records are append-only.
8. Zero external dependencies. All enforcement scripts use Python stdlib only.
| Signal | Mode | Workflow |
|---|---|---|
| "audit", "analyze", "document", "snapshot", "index the repo" | AUDIT | workflows/audit.md |
| "amend", "change", "fix", "add", "update", names files to change | AMEND | workflows/amend.md |
If unclear, ask: "Do you want to audit the repo (read-only analysis) or amend it (make a code change with cross-cutting documentation)?"
After reading the workflow, follow it exactly.
All in workflows/:
| Workflow | Purpose |
|---|---|
| audit.md | Full repo audit β setup enforcement, inventory, scaffold, generate reports, patterns, snapshot |
| amend.md | Code change with cross-cutting documentation and amendment record |
All in references/:
| Reference | Purpose |
|---|---|
| enforcement-spec.md | How enforcement works β pre-commit hook, CI check, exclusions |
| amendment-fields.md | Required and recommended YAML fields for amendment records |
All in templates/:
| Template | Purpose |
|---|---|
| amendment-record.md | Full YAML + markdown structure for amendment records |
Bundled enforcement files in scripts/ (relative to this SKILL.md). These are portable templates β copied to target repos during audit setup.
| Script | Destination in target repo |
|---|---|
| _audit_config.py | scripts/_audit_config.py |
| repo_audit.py | scripts/repo_audit.py |
| enforce_amendment_protocol.py | scripts/enforce_amendment_protocol.py |
| pre-commit-config.yaml | .pre-commit-config.yaml |
| enforce-amendment.yml | .github/workflows/enforce-amendment.yml |
The repo-audit skill succeeds when:
- AUDIT mode: all artifacts are generated (per-file reports, patterns.md, CLAUDE.md, supporting files) and enforcement is installed
- AMEND mode: a valid amendment record is committed alongside the code change, with all required fields populated and pre-commit/CI validation passing
# Related Skills
Use when feature flag tests fail, flags need updating, understanding @gate pragmas, debugging...
Use when you have lint errors, formatting issues, or before committing code to ensure it passes CI.
Use when you need to check feature flag states, compare channels, or debug why a feature behaves...
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Amp
Antigravity
Claude Code
Clawdbot
Codex
Cursor
Droid
Gemini CLI
GitHub Copilot
Goose
Kilo Code
Kiro CLI
OpenCode
Roo Code
Trae
Windsurf
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.