Anvil

Audit AI rule files in any codebase. Score coverage gaps, detect drift, mine PR history for missing rules, and bootstrap new rule sets from your tech stack.

Start with one real audit

bunx @lambdacurry/anvil audit \
  --target . \
  --ci

This local-only lane gives new users a real report on the first try with just Bun. If you are one directory above the target repo instead, swap --target . for --target ./my-repo. For the full AI-backed lane or alternate launchers, continue to Installation or First Audit.

Audit rules

Score existing rules for coverage gaps, conflicts, drift, and format compliance.

Detect drift

Flag stale paths, broken globs, outdated validation dates, broken symlinks, and command drift. Coverage analysis is planned.

Bootstrap rules

Generate starter rule sets from your tech stack analysis.

Mine PRs

Surface recurring review feedback that should become rules.

What Anvil checks

• Coverage gaps — patterns with no rule protection in the audit; drift-side coverage analysis is planned
• Drift — path, glob, date, broken-symlink, and command checks are implemented
• Conflicts — contradictory instructions across rule files
• Sizing — rules that bloat context without earning their place
• Format — compliance with rule-writing best practices

Framework-agnostic. Works on any AI-assisted codebase.

Found something?

Anvil is in alpha and looking for real-world feedback. If a command fails, a score looks wrong, or the docs are unclear, open a GitHub issue — first-run reports are especially valuable.