Code Review deux axes (Matt Pocock skill)
You invoke /code-review on the diff since a fixed point (a SHA, a branch, a tag, main) and the skill spawns two parallel subagents: Standards (does the code follow the repo's documented conventions?) and Spec (does the code do what the originating issue asked?). It first checks the ref resolves and the diff is non-empty, then reports both verdicts side by side, never merged or reranked. The Standards axis carries a hardcoded baseline of twelve Fowler code smells (Refactoring, ch.3), applied even when the repo documents nothing.
Strengths
- Two parallel subagents with separate contexts, so the Spec axis does not get contaminated by style remarks
- Twelve Fowler code smells baked in: the review still holds on a repo that documents no convention at all
- A documented repo standard always overrides the baseline, and each smell is labelled a judgement call, never a hard violation
- Fail-fast on the ref: a bad ref or empty diff stops before burning two subagents
Limitations
- The Spec axis is useless without a spec: with no issue and no spec file the subagent is skipped and half the review is gone
- Assumes the tracker is configured (docs/agents/issue-tracker.md), so /setup-matt-pocock-skills is a de facto prerequisite
- Does not hunt bugs: it is a conventions plus conformance review, no substitute for a correctness pass or the tests
Best for
- Devs who want a review before pushing a branch, without waiting on a human
- Teams with a CONTRIBUTING.md or standards file nobody ever rereads at diff time
- Repos where the gap between what the ticket asked and what got coded regularly ships