code-review
Review a branch, PR, work in progress, or changes since a ref against repository standards and the originating requirements.
- License 许可
- MIT
- 最近更新
- 2026-11-12
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- code-review
- description
- Review a branch, PR, work in progress, or changes since a ref against repository standards and the originating requirements.
02 资源文件 3
- 许可证LICENSE1.0 KB
- 参考文档references/review-surface.md2.0 KB
- 参考文档references/smell-baseline.md1.2 KB
03 SKILL.md 正文
Review a captured change on two independent axes: Standards (repository rules) and Spec (the originating requirements). Keep their findings separate.
1. Capture the requested scope
Use the requested base and scope, including any already established in the task. For a WIP-only review with no base, use HEAD. For a branch or PR, use its known target; if the target is unknown, ask. Resolve base, HEAD and merge-base to commit IDs.
Follow review-surface.md to capture the review packet. A branch/PR review covers committed changes unless pending work was requested; an implementation review includes the pending work being implemented. State exclusions. An invalid ref stops capture. An empty committed diff alone does not stop a WIP review. If every requested layer is empty, report no changes without launching reviewers or claiming a review pass.
2. Find requirements and standards
Read docs/agents/issue-tracker.md when resolving tracker artifacts. If missing,
ask the user to run /setup-matt-pocock-skills. Find requirements in this order:
- The approved spec or maintenance plan already supplied in the task.
- A path or issue reference supplied by the user.
- Issue references in captured commit messages, using the configured tracker.
- A matching spec under the configured tracker paths.
- If none is found, ask where the requirements are. If the user confirms there are none, skip Spec and report "no spec available"; do not claim compliance.
Find applicable repository standards such as AGENTS.md, CONTRIBUTING.md and CODING_STANDARDS.md. For the Standards axis, also load the smell baseline. These are labeled heuristics; documented repository standards override them. Skip checks already enforced by tooling.
3. Run independent reviews
Run Standards and Spec as parallel read-only subagents: the Agent tool with
harness-reviewer, one axis each.
Give both the same captured diffs, relevant file contents, scope manifest,
base/HEAD IDs and commit list. A diff command alone is not a snapshot.
Keep the checkout stable while they read surrounding context.
Standards assignment: Include applicable standards and the loaded smell baseline. Report each documented-standard breach with file:line, the rule and evidence. Report smells separately as judgment calls, naming the smell and quoting the relevant hunk. Explain the impact; repository rules override the baseline.
Spec assignment: Include the requirements. Report missing or partial behavior, unrequested scope, and incorrect implementations. Cite the requirement and the captured file:line for each finding, explaining the impact.
Ask each reviewer to report every finding, including uncertain or low-severity ones, with its evidence, confidence and severity (high, medium or low impact). At this stage the job is coverage; the user decides what to act on. If Spec has no source, run only Standards. If independent contexts are unavailable, use isolated passes and disclose that limit.
4. Report
Check the snapshot for drift as described in review-surface.md. Present findings
under ## Standards and ## Spec, verbatim or lightly cleaned; do not merge or
rerank them across axes. End with finding counts and the worst issue within each
axis. State the reviewed scope, skipped coverage and evidence limits.
A change can follow standards while implementing the wrong requirement, or match requirements while violating standards. One axis passing does not imply the other.
04 同仓库更多 Skills
writing-for-agents
Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.
wizard
Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.
wayfinder
Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.
wait-what
Stop. That last message did not land: re-pitch it.
to-tickets
Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker (edges as text in one file per ticket locally, or native blocking links on a real tracker).
to-spec
Turn the current conversation into a spec and publish it to the project issue tracker: no interview, just synthesis of what you've already discussed.