code-review

/code-review — a model-invoked engineering skill that reviews the diff since a fixed point (commit, branch, tag, or merge-base) along two axes, run as parallel sub-agents so neither pollutes the other’s context, then aggregated:

• Standards — does the code follow the repo’s documented coding standards, plus an always-on Fowler smell baseline (~12 curated “Bad Smells in Code”: Mysterious Name, Duplicated Code, Feature Envy, Data Clumps, Primitive Obsession, Repeated Switches, Shotgun Surgery, Divergent Change, Speculative Generality, Message Chains, Middle Man, Refused Bequest). Two binding rules keep it safe: a documented repo standard overrides the baseline, and every smell is a judgement call, never a hard violation.
• Spec — does the code faithfully implement the originating issue / spec (found via commit refs, a passed path, or a docs//specs/ file)?

The sub-agent dispatch is written harness-neutrally, so the step works on Codex as well as Claude Code. Refactoring belongs to this skill rather than to mp-tdd.

In the main flow it is the quality gate after the build: once per ticket under mp-implement, or once over the whole integration branch under mp-implement-spec. mp-retro then tunes the coding standards it enforces, turning mechanical findings into deterministic checks and leaving CODING_STANDARDS.md for judgement calls.

MP’s member of the cross-framework code-review cluster (pattern-adversarial-review smell- hunting via pattern-parallel-persona-review sub-agents), with addy-code-review, bmad-code-review, ce-code-review, and gstack-review (a five-framework cluster). Its distinctive shape is the explicit Standards ∥ Spec two-axis split.

See Also

• mp-implement — invokes this to close out a build.
• mp-retro — tunes the standards this enforces.
• mp-improve-codebase-architecture — MP’s other stage-review skill (architecture audit, not diff review).
• addy-code-review · bmad-code-review · ce-code-review · gstack-review — code-review-cluster counterparts.
• stage-review — the canonical stage this implements.

Version history

v1.3.0 (2026-10-04)

• Runs once per ticket under mp-implement or once over the integration branch under mp-implement-spec; mp-retro added downstream to tune its standards.

v1.2.0 – v1.2.3 (2026-08-05 – 2026-08-06)

• The spec axis talks about “issue/spec” rather than “issue/PRD” (v1.2.0).
• Claude Code’s tool and agent-type names dropped from the sub-agent dispatch, so it works on Codex too (v1.2.3).

Added in v1.1.0 (2026-07-08; ingested 2026-07-09)

• Graduated from in-progress/ (was review) into engineering/. The refactor stage of mp-tdd moved here, and it became the quality gate mp-implement runs before committing.