bmad-code-review

bmad-code-review — “Review code changes adversarially using parallel review layers (Blind Hunter, Edge Case Hunter, Acceptance Auditor) with structured triage.” Owned by bmad-dev and run in “fresh context, ideally different LLM” than the one that wrote the code (pattern-fresh-context-subagents) → artifact-review-report.

It combines two patterns: it fans several single-lens reviewer layers out in parallel (pattern-parallel-persona-review), each under the adversarial mandate that “the reviewer must find issues. No ‘looks good’ allowed” — with the honest caveat that “because the AI is instructed to find problems, it will find problems … Human filtering remains the essential final step.”

Cross-framework cluster (code review)

• addy-code-review — Addy’s five-axis review with severity labels + change sizing.

Both produce a artifact-review-report; BMAD’s distinction is the multi-layer adversarial fan-out over a single diff, versus Addy’s structured single-pass rubric.

See Also

• bmad — the framework.
• addy-code-reviewer, gsd-verifier — related quality/verification agents.
• The Test Architect (TEA) module (Murat) adds enterprise quality gates BMM core lacks — see bmad.
• stage-review — the canonical stage.