code-review
Skill0 starsChanged 46 days agoMITView on GitHub
Reviewing a diff or PR: severity triage, what to flag vs let go, actionable findings, merge verdict. Used by /review and the code-reviewer agent.
194 other projects use a version of this.
Adopted by
Most recently taken up by
- dzhng/skillstheir own version, first seen Nov 15, 2026
- mdiri-ai/SoduWorktheir own version, first seen Nov 15, 2026
- sirkirby/unifi-mcptheir own version, first seen Nov 13, 2026
- microsoft/vscodetheir own version, first seen Nov 12, 2026
- oceanbase/seekdbtheir own version, first seen Nov 12, 2026
- dzhng/jevgreptheir own version, first seen Nov 11, 2026
Used with
Projects with it also use close-spec, design-with-images 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
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
PydanticSkilladd-new-model
Add support for a newly-released language or image generation model in pydantic-ai (e.g. openai:gpt-5.6, anthropic:claude-sonnet-5, openai:gpt-image-2). Use when a provider ships a new model id and you need to wire literals, profile flags, adapters, and tests to recognize it. Handles SDK-lag, gateway list conventions, capability probing, and direct image-model geometry.
Also in 9 projects that use this
transitive-bullshitSkillcompare-screenshots
Compare screenshots against the intended design, distinguishing approved references from historical baselines. Use for iterative implementation matching, before/after visual review, objective image telemetry, or checking a lone capture for flat, empty, or badly framed content.
Also in 5 projects that use this
Microsoft AzureSkillazsdk-common-apiview-feedback-resolution
Analyze and resolve APIView review feedback on Azure SDK PRs. **UTILITY SKILL**. USE FOR: APIView comments, API review feedback, SDK API surface changes. DO NOT USE FOR: general code review, non-APIView feedback. INVOKES: azure-sdk-mcp:azsdk_apiview_get_comments, azure-sdk-mcp:azsdk_customized_code_update, azure-sdk-mcp:azsdk_typespec_delegate_apiview_feedback, azure-sdk-mcp:azsdk_run_typespec_validation, azure-sdk-mcp:azsdk_package_generate_code.
Also in 6 projects that use this
Microsoft AzureSkillazsdk-common-generate-sdk-locally
Generate and finalize Azure SDKs locally from TypeSpec, including build, checks, tests, changelog, metadata, and version updates by default. WHEN: "generate SDK locally", "build SDK", "run SDK tests", "run CI checks", "validate package", "run checks", "update changelog", "fix SDK build errors", "resolve SDK generation errors", "customize TypeSpec", "rename SDK client", "rename SDK model", "hide operation from SDK", "fix analyzer errors", "resolve customization drift", "create subclient", "update metadata", "update version". INVOKES: azsdk_verify_setup, azsdk_package_generate_code, azsdk_package_build_code, azsdk_package_run_check, azsdk_package_run_tests, azsdk_customized_code_update, azsdk_package_update_changelog_content, azsdk_package_update_metadata, azsdk_package_update_version.
Also in 6 projects that use this
- +4 more from dzhng
- +2 more from Azure
Other versions194 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 194
- mattpocock/skills★ 325ksame name
- TheAlgorithms/Python★ 226ksame name
- microsoft/vscode★ 195ksame name
- github/spec-kit★ 150ksame name
- ggml-org/llama.cpp★ 135ksame name
- supabase/supabase★ 115ksame name
- storybookjs/storybook★ 92ksame name
- shareAI-lab/learn-claude-code★ 80ksame name
- shareAI-lab/analysis_claude_code★ 80ksame name
- vectorize-io/hindsight★ 61ksame name
184 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.
In this repository29 other files it gives its agents.
- hosts/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- adrSkill
- api-designSkill
- auditSkill
- concurrency-performanceSkill
- coverageSkill
- debuggingSkill
- debugSkill
- fast-laneSkill
- fixSkill
- implementSkill
- intakeSkill
- leanSkill
- migration-safetySkill
- nonnaSkill
- observabilitySkill
- planSkill
- refactoringSkill
- releaseSkill
- reviewSkill
- rollbackSkill
- security-reviewSkill
- shipSkill
- supply-chainSkill
- syncSkill
- tddSkill
- tdd-workflowSkill
- testSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 183 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 29
Secrets & logging: No keys/tokens/PII in code, logs, or traces.
- Distinctive: line 30
Reproducibility: No wall-clock or global RNG in decision logic.
- Distinctive: line 64
Prefer fewer, high-confidence findings.
“Common” is measured: kept by at least 6 in 10 of the 183 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
Code Review
Review as an adversary, report as an ally. Your job is to find the defect that reaches production, not to repaint the bikeshed. Safety is lexicographically prior to speed — a fast PR that fails open is a defect, not a feature. The reviewer proposes findings; a deterministic gate decides the merge (see "Verdict" below).
Severity, in one breath
CRITICAL = trust-boundary breach, silent failure in money/data/state, or fail-open — blocks merge,
no debate. HIGH = a correctness bug on a real path, missing tests on critical logic, a swallowed
error, or a secret/PII in logs — blocks merge. MEDIUM = weak handling or an edge case off the hot
path — fix when feasible. LOW = readability/naming — optional. Full calibration, the auto-CRITICAL
list, and how severity maps to the gate: references/severity-rubric.md.
Category checklist
- Correctness & boundaries. Does it do what the change claims? Is untrusted input validated at
the edge before it is trusted? Are illegal states reachable? (
.claude/rules/boundaries.md) - Tests present & meaningful. New logic has tests that would fail without the change. Check the asserts are real, not vacuous. No test → treat as incomplete.
- Error handling / fail-closed. On error, does it deny, halt, or roll back — never proceed? Are
errors typed and propagated, not swallowed? (
.claude/rules/safety.md) - Secrets & logging. No keys/tokens/PII in code, logs, or traces. No secret in a fixture.
- Reproducibility. No wall-clock or global RNG in decision logic; same inputs → same outputs.
- Style fit. Matches the local idiom. Consistency with the file beats your personal preference.
- Complexity. The ladder (
.claude/rules/00-core.md) applied to the diff: hand-rolled stdlib, a new dependency for what the platform does, an abstraction with one implementation, adebt:marker with no upgrade trigger. Tag theissuedelete:/stdlib:/native:/yagni:/shrink:, name the replacement infix,category: simplicity, MEDIUM at most — size never outranks correctness or security. When any exist, end the prose withnet: -N lines possible.Depth and examples: theleanskill.
Reviewing as an adversary
Assume the diff is hiding a bug and go find it. Trace the untrusted path: where does external input enter, and what is the first place it is trusted without a check? Ask the failure questions — what happens when this is null / negative / huge / concurrent / retried / times out? Read the test the author didn't write. Confirm fail-closed on the error branch, not just the happy path.
Writing a finding
One finding = path:line + the problem in one sentence + a concrete fix. Make it cheap to act on.
payments/refund.py:88— CRITICAL: refund amount comes straight from the request body with no bound check; a negative or oversized value posts a fraudulent credit. Validate0 < amount <= original_chargeat the handler edge and reject otherwise.
Bad finding: "error handling could be improved here." No location, no defect, no fix — noise.
A finding whose fix adds code states the failing input in issue; without one it is a
question, not a MEDIUM.
What NOT to flag
Signal dies in a flood of nits. Skip subjective style a formatter/linter already owns, restyling that fights the local idiom, speculative hardening (RFC-grade validation, limits, i18n cases) on a surface the task did not make critical — and never ask for speculative "what if we later need…" generality; when the diff adds it, that is a Complexity finding. Prefer fewer, high-confidence findings. If unsure it's a real defect, mark it as a question, not a blocker — ten weak findings bury the one that matters.
Verdict — emit JSON, let the gate decide
Close by emitting the structured verdict — schema:
templates/verdict.json. Each finding is one object
(severity, path, line, category, issue, fix); the top-level verdict is
approve or request_changes.
Pipe it through the deterministic gate, which exits 1 on request_changes or any CRITICAL/HIGH:
your-review-step | .claude/skills/code-review/scripts/check-review.sh
# or: .claude/skills/code-review/scripts/check-review.sh review.jsonThis is the boundary doing its job (.claude/rules/boundaries.md): the LLM proposes severities; the
script renders the merge decision. Safe-to-merge once every CRITICAL and HIGH is addressed;
MEDIUM/LOW may ship with a follow-up. Hand security-sensitive diffs to the security-reviewer.
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.

