agent-config-auditlisted

Audit AI agent configuration files (CLAUDE.md, AGENTS.md, preference files) for contradictions, duplication, bloat, and personal/team boundary violations. Use when unsure which instruction files are active for a project, after restructuring how your config is organised, or when the same rule turns up in two files with different wording. Do NOT fire to write or restructure one file (that is `claude-md-pointer-check`), to judge whether a single addition earns its place (`redundancy-check-before-ship`), or on source, tests and config values: this reads the instruction files as a set and reports what they do to each other.
damson/hard-won-skills · ★ 1 · AI & Automation · score 67

Install: claude install-skill damson/hard-won-skills

# Agent Config Audit Audit the AI instruction files active for the current project. Report findings, optionally fix them. ## Step 1: Resolve the Active File Stack From the current working directory, resolve the project's instruction-file stack — items 1, 3 and 4 are loaded by the harness; 2 and 5 are audited when present (all five resolved; scoping flags filter afterwards — see Flags): 1. **Global CLAUDE.md** — `~/.claude/CLAUDE.md` 2. **Global preference/rule files** — whatever else `~/.claude/` loads. Enumerate, don't assume names: `ls ~/.claude/*.md ~/.claude/rules/*.md 2>/dev/null`, minus the `CLAUDE.md` item 1 already covers — take what exists 3. **Project AGENTS.md** — `AGENTS.md` in the project root 4. **Project CLAUDE.md** — `CLAUDE.md` in the project root 5. **Project preference/rule files** — under the project's `.claude/` Any of these may be a symlink into a config repo — follow the chain to its source (`readlink -f`) so the audit edits the real file, not a link For each file found, report: - The logical path (e.g. `<project>/AGENTS.md`) - The resolved source path, where it differs (e.g. `<config-repo>/<domain>/AGENTS.md`) - Whether it is personal or team-owned. The rule: a file that resolves into the user's config repo or lives under `~/.claude/` is personal; a file tracked directly in the project repo is team-owned Print the stack before running checks so the user knows the scope. ## Step 2: Run Checks For each finding, record: severity (`🔴