bmad-dev-auto

Invoke the skill with a clear intent such as a story ID, ticket number, or short feature description. It autonomously clarifies the intent, writes a spec, implements the code via subagents, reviews the diff, and commits the result without asking questions.

Step 4 launches two parallel review subagents, Blind Hunter and Edge Case Hunter, on the diff since the baseline revision. Findings are deduplicated, severity-rated, and triaged into intent_gap, bad_spec, patch, defer, or reject categories with automatic repair loops.

Yes. Override defaults in _bmad/custom/bmad-dev-auto.toml for team settings or bmad-dev-auto.user.toml for personal settings. You can add activation steps, persistent facts, and an on_complete instruction; scalars override and tables deep-merge.

It executes the HALT protocol: updates the spec file status, writes a result file with the blocking condition, runs any configured on_complete instruction, and stops. Blocking conditions include unclear intent, missing subagents, or exceeding 5 review repair loops.

Yes. When the intent is an epic story, it compiles or loads a cached epic context file distilled from planning artifacts, and pulls continuity context from the most recent completed story spec in the same epic.