review
Skill102 starsChanged 35 days agoMITView on GitHub
After building a feature, verify it matches what was planned, respects the system architecture and design standards, and is ready for production. Reports issues clearly so the developer decides what to fix.
192 other projects use a version of this.
Adopted by
Most recently taken up by
- dzhng/skillstheir own version, first seen Nov 15, 2026
- muxy-app/muxytheir own version, first seen Nov 15, 2026
- macro-inc/macrotheir own version, first seen Nov 15, 2026
- igoroliveirg/claude-copytheir own version, first seen Nov 15, 2026
- yagizdo/quivertheir own version, first seen Nov 15, 2026
- markmhendrickson/neotomatheir own version, first seen Nov 15, 2026
Used with
Projects with it also use close-spec, daily-arxiv and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
dzhngSkillclose-spec
When a spec under specs/<feature>/ is done shipping (a write-spec build, or any planned task that produced a spec), archive it to specs/done/ and rewrite it from a build-plan into a durable rationale — the why, the principles, the invariants — pointing back to the real code for the how. Use when implementation has landed and the plan no longer matches what shipped, or the user says a feature/spec is finished. Pairs with [write-spec](../write-spec/SKILL.md) (the plan this closes) and [review](../review/SKILL.md) (run before closing).
Also in 5 projects that use this
LambenthanSkilldaily-arxiv
每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新
Also in 5 projects that use this
dzhngSkilldesign-with-images
Explore UI and visual design with image generation, then iterate the real implementation against the user-selected concept. Use when brainstorming visual directions, offering mockup variants, or implementing an approved generated design; do not stop at the mockups or the first code approximation.
Also in 5 projects that use this
LambenthanSkillexp-eval
实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges
Also in 5 projects that use this
- +5 more from Lambenthan
- +3 more from dzhng
Other versions192 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 192
- garrytan/gstack★ 142ksame name
- ansible/ansible★ 71ksame name
- docling-project/docling★ 71ksame name
- traefik/traefik★ 66ksame name
- ClickHouse/ClickHouse★ 51ksame name
- temporalio/temporal★ 24ksame name
- ccch1mneyyy/dsh-Go★ 24ksame name
- tisfeng/Easydict★ 15ksame name
- juicedata/juicefs★ 15ksame name
- elie222/inbox-zero★ 13ksame name
182 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 11 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 156 edited versions we read keep these.
11
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Run this after every feature.
- Distinctive: line 20
Before running this skill:
- Distinctive: line 36
Before reviewing anything, establish the benchmark.
- Distinctive: line 38
Read in this order:
- Distinctive: line 44
If no plan exists, ask the developer to describe what the feature was supposed to do before reviewing.
- Distinctive: line 56
Every part of the feature description — is it all there?
- Distinctive: line 86
After completing all three layers, produce a clear report.
- Distinctive: line 114
After presenting the report, stop.
- Distinctive: line 116
Wait for the developer to:
- Distinctive: line 118
Ask you to fix a specific issue
- Distinctive: line 128
Not all issues are equal. Use this to help the developer prioritise:
“Common” is measured: kept by at least 6 in 10 of the 156 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Building is not done when the code runs. It is done when the code is correct.
AI moves fast. Fast means things get built that work on the surface but drift from the architecture, violate the design system, or miss edge cases that matter. This skill catches those things before they compound into bigger problems.
Run this after every feature. Before you move on.
Critical Requirement — Fresh Context Only
This skill must always run in a fresh context window — not in the same session that built the feature.
The agent that built the feature carries the assumptions, shortcuts, and blind spots of the build session. It will review its own work and find it correct — not because it is, but because it cannot see what it has already decided to overlook.
Before running this skill:
- End the build session with
/remember save - Open a new context window
- Let
/remember restorerun automatically - Then run
/review
A self-review is not a review. It is a confirmation bias check that confirms the bias.
What This Skill Does Not Do
It does not fix anything. It reports what it finds and lets the developer decide what matters and what to do about it. Fixing without understanding is how problems get buried, not solved.
Step 1 — Understand What Should Have Been Built
Before reviewing anything, establish the benchmark.
Read in this order:
- The implementation plan from
/architectif one exists - The feature description or task that was given
- Any relevant context files — architecture boundaries, code standards, design rules
If no plan exists, ask the developer to describe what the feature was supposed to do before reviewing. You cannot verify correctness without knowing what correct looks like.
Step 2 — Review in Three Layers
Layer 1 — Does it match the plan?
Compare what was built against what was planned.
Check:
- Every part of the feature description — is it all there?
- The decisions made during planning — are they reflected in the code?
- The scope — did the implementation stay within bounds or add things that were not asked for?
Flag anything that was planned but missing. Flag anything that was built but not planned.
Layer 2 — Does it respect the system?
This is where AI drift most commonly happens. The feature works, but it violates rules that the project depends on.
Check:
- Architecture boundaries — does code in the right place own the right responsibilities? No UI logic in API routes. No DB calls in components. Whatever the project's boundaries are — are they respected?
- Design system — are the correct tokens, classes, and patterns used? Any hardcoded values that should be variables? Any raw color classes that should use the design system?
- Code standards — naming conventions, file organisation, TypeScript strictness, error handling patterns — do they match what the project established?
- Existing patterns — does this feature introduce a new pattern when an existing one should have been used?
Layer 3 — Is it production ready?
Check:
- Error handling — what happens when things go wrong? Are errors caught and handled or does the feature silently fail?
- Edge cases — empty states, loading states, missing data — are these handled?
- Console errors — any errors or warnings in the browser or terminal?
- Obvious bugs — anything that would clearly break for a real user?
Step 3 — Report What You Found
After completing all three layers, produce a clear report. Do not bury issues. Do not soften them. Report honestly so the developer can make informed decisions.
## Review — [Feature Name]
### Layer 1 — Plan alignment
[PASS / ISSUES FOUND]
[List any gaps between what was planned and what was built]
### Layer 2 — System integrity
[PASS / ISSUES FOUND]
[List any architecture, design, or code standard violations]
### Layer 3 — Production readiness
[PASS / ISSUES FOUND]
[List any error handling gaps, edge cases, or obvious bugs]
### Summary
[X] issues found across [Y] layers.
[If no issues: "No issues found. This feature is ready to ship."]
[If issues: "Resolve the above before moving to the next feature."]Step 4 — Let the Developer Decide
After presenting the report, stop. Do not start fixing. Do not suggest fixes unless the developer asks.
Wait for the developer to:
- Ask you to fix a specific issue
- Tell you an issue is intentional and can be ignored
- Confirm everything is resolved and ready to move on
The developer owns the quality decision. You inform it.
Severity Guide
Not all issues are equal. Use this to help the developer prioritise:
Critical — fix before moving on
- Architecture boundary violations that will break future features
- Missing error handling that causes silent failures
- Functionality that was planned but completely missing
Important — fix soon
- Design system drift that will cause UI inconsistency
- Code standard violations that will compound across the codebase
- Edge cases that a real user will encounter
Minor — fix when convenient
- Naming inconsistencies that do not affect behaviour
- Missing optimisations
- Style issues that do not affect the design system
Label each issue with its severity so the developer can triage quickly.
The Standard
The question this skill answers is not "does it work?"
The question is "is it correct?"
Working and correct are not the same thing. A feature can work today and break the project tomorrow. Review exists to catch the difference.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

