code-review

This skill (a Chinese-language translation of Matt Pocock's skills) performs a two-axis code review of the diff between a user-supplied fixed point (commit, branch, tag, or merge-base) and HEAD. The two axes are Standards (does the code follow the repository's documented coding standards?) and Spec (does the code faithfully implement the source issue/PRD/spec?). It solves the problem of one concern masking another: a change can pass standards while implementing the wrong thing, or match the issue exactly while breaking project conventions, so the two reviews run in parallel sub-agents and are reported side by side without merging.

The workflow pins the fixed point (validating that the ref resolves and the diff is non-empty via a three-dot `git diff` and `git log`), identifies the spec source from commit issue references, a passed path, or PRD files, and identifies standards sources such as CODING_STANDARDS.md or CONTRIBUTING.md. Beyond documented standards, the Standards axis always applies a fixed baseline of Fowler code smells (mysterious name, duplicated code, feature envy, data clumps, primitive obsession, repeated switches, shotgun surgery, divergent change, speculative generality, message chains, middle man, refused bequest), with the rule that documented repo standards override the baseline and that smells are judgement calls rather than hard violations. Two general-purpose sub-agents run with focused briefs, and the skill aggregates their findings under separate headings.

Target users are developers and AI agents who want to review a branch, PR, or in-progress change, or run a "review since X." The skill only reads diffs and specs and spawns review sub-agents; it makes no changes to the code.