bmad-project-context

Invoke the skill with the setup intent against your repository. It scans configs, CI, and source for verifiable facts, interviews you for policies and conventions no scan can find, then writes a small approved block between markers in AGENTS.md.

Include only what agents cannot derive from the repo: policy like branch rules and frozen paths, caveats about running tests, conventions differing from ecosystem defaults, and observed pitfalls. Exclude directory trees, stack lists, and commands already stated in package.json or CI config.

Use the record intent when an agent makes a mistake. The skill captures the task, mistake, correction, and evidence, then adds a pitfall line or exact command correction to the block, proposing a hook or CI check if the error is mechanically preventable.

No. Writes are spliced only between the block markers, leaving everything outside byte-identical, and the user approves the complete block before any write. It never commits changes to the repository.

Refresh after significant repository changes; the skill diffs deletions and renames since the recorded commit SHA and re-verifies every line. Audit re-checks caveats and removes lines whose removal would not change agent behavior.