What problem does it solve? Repositories accumulate stale, bloated, or missing agent instruction files, causing AI coding agents to repeat mistakes, run wrong commands, and violate team policies. This Skill produces a small, verified, managed block inside AGENTS.md so agents work correctly in that repo. ## Core Features & Use Cases - Setup, adopt, refresh, record, and audit intents: Scaffold new instructions, migrate existing CLAUDE.md or AGENTS.md content into a managed block, re-verify against the current codebase, capture observed agent mistakes as pitfalls, and prune dead lines. - Evidence-based verification: Scans package.json, Makefiles, pyproject.toml, and CI config, path-checks every claim, and diffs against a recorded commit SHA so the block never drifts from reality. - Ledger-driven migration: Every existing instruction is tracked through retain, rewrite, relocate, automate, or delete decisions with explicit grounds, so nothing is silently lost. - Use Case: A team adopting AI coding agents runs the adopt intent to consolidate a sprawling CLAUDE.md into a concise AGENTS.md block, with each original instruction accounted for and unverifiable claims removed. ## Quick Start Ask the agent to set up or refresh this repository's AGENTS.md agent instructions using the bmad-project-context skill.
