What problem does it solve? Project-rules files like AGENTS.md and CLAUDE.md often become stale, incomplete, or unsupported by the AI client in use, causing agents to work with inaccurate project context. This Skill audits every rules file in a repository, grades each against a quality rubric, and applies targeted improvements only after user approval. ## Core Features & Use Cases - Full rules-file discovery: Scans the repository for AGENTS.md, CLAUDE.md, CLAUDE.local.md, .claude/rules files, OpenCode instructions, and Claude @ imports, with client-specific resolution behavior for Codex, OpenCode, and Claude Code. - Graded quality assessment: Scores each file on commands, architecture clarity, currency, and actionability, producing an A–F grade with a per-criterion breakdown. - Approval-gated updates: Presents a quality report first, then proposes minimal diffs and applies edits only after user confirmation. - Use Case: After a major refactor, ask the agent to audit your AGENTS.md; it finds stale build commands and missing test instructions, shows a scored report, and updates the file with your approval. ## Quick Start Ask the agent to audit and improve the AGENTS.md and CLAUDE.md files in this repository and show a quality report before making any changes.
