bmad-project-context

Run the skill's setup intent against your repository. It scans config files, CI, and source, interviews you only about what scans cannot answer, then shows the complete block for approval before splicing it between markers in AGENTS.md.

Only what an agent cannot derive by reading the repo: policy the code cannot express, caveats about running the project, conventions differing from ecosystem defaults, and pitfalls with observed evidence. Repo overviews, style rules, and commands already in package.json or CI config are excluded.

Use the record intent with the task, the mistake, the correction, and its evidence. One occurrence is noted; a recurring or costly mistake earns a pitfall line or an exact command line, and mechanically preventable issues are routed to a hook, lint rule, or CI check instead.

No. The skill is conversational and never writes until it shows the complete block and receives your approval. It splices content only between markers, leaves everything outside byte-identical, and never commits.

Refresh reads the provenance line's recorded SHA, re-verifies every path and caveat, and runs git log for deletions and renames since that SHA against every line. Lines whose evidence is gone are updated or removed, and the interview shrinks to what changed.

A component or nested repository gets its own file when work keeps landing there and its rules do not belong at the parent level. Rules bounded to a directory go in a nested AGENTS.md attached by location, and every child is listed in the parent's Where things are section.