Skill Finder
The best Claude skills for code review
For most people, start with code-review-and-quality from Addy Osmani: it reviews any change on five axes, from correctness to performance, and needs nothing extra. react-doctor scores highest, 78, but only checks React code, and review-animations, at 77, only judges motion code. The list also covers ponytail-review and ponytail-audit, rules for Rust, Go and SwiftUI, CI review output, and secondary picks.
The 14 skills made for this job
-
1
Get review-animationsreview-animations
For front-end teams who want a strict second opinion on animation code before it ships. Skip it if you need a general code review or help writing new motion.
-
2
Get code-review-and-qualitycode-review-and-quality
For developers who want a structured review of a diff or pull request before merging. Skip it if you only want a quick style lint or the project has no tests to run.
-
3
Get react-doctorreact-doctor
For React teams who want a health score and regression check before each commit. Skip it if you do not want Claude running npx packages or following a playbook fetched from the web.
-
4
Get ponytail-reviewponytail-review
For developers who want a plain-English review of a change that ends in a ship or fix verdict. Skip it if you want the fixes applied, since it only reports.
-
5
Get rust-best-practicesrust-best-practices
For Rust developers who want code written and reviewed against Apollo GraphQL's handbook. Skip it if you are not writing Rust or your team already follows its own style guide.
-
6 Get ponytail-audit
ponytail-audit
For developers or founders who want a whole repo audited in plain English and ranked by risk. Skip it if you want code changed or one diff reviewed.
-
7
Get swiftui-proswiftui-pro
For iOS and Mac developers who want SwiftUI code reviewed for outdated APIs, accessibility and performance. Skip it if your app is built mainly in UIKit or not in Swift at all.
-
8
Get thermo-nuclear-code-quality-reviewthermo-nuclear-code-quality-review
For developers who want a very strict maintainability review of a branch that pushes for big simplifications. Skip it if you want a quick, gentle review or a bug and security hunt.
-
9
Get ast-grepast-grep
For developers who need to find code by its structure, like async functions without try-catch. Skip it if plain text search already finds what you need.
-
10
Get golang-code-stylegolang-code-style
For Go developers who want clarity rules a linter cannot enforce, such as early returns and named conditions. Skip it if you want naming or linter setup, which sibling skills cover.
-
11
Get review-prreview-pr
For teams running automated PR reviews in CI who want structured JSON output instead of direct GitHub posts. Skip it if you want Claude to comment on a PR by hand.
-
12
Get better-interfacebetter-interface
For teams who want one ranked UI review covering accessibility, layout, copy, type, color and polish. Skip it if you lack the other better-* skills it hands each area to.
-
13
Get karpathy-guidelineskarpathy-guidelines
For developers who want Claude to ask before guessing, keep changes small and prove work with tests. Skip it for quick throwaway scripts where the caution slows you down.
-
14
Get simplifysimplify
For TypeScript and React developers who want freshly written code tidied without behavior changes. Skip it if your style rules differ, since it ships with one author's preferences.
Also useful for code review
Their main job is something else, but they help with this one.
-
Get vercel-react-best-practicesvercel-react-best-practices
For React and Next.js developers who want slow pages and heavy bundles fixed by a ranked rulebook. Skip it if you do not write React or Next.js code.
-
Get CavemanCaveman
For developers who want terse commit messages, one-line review comments and lower token spend. Skip it if you prefer full, explanatory prose from Claude.
-
Get security-auditsecurity-audit
For developers who want a source-based security review, from one question up to a full multi-agent audit. Skip it if you want scans of live deployed sites.
-
Get emil-design-eng
emil-design-eng
For designers and front-end developers who want tuned motion and small touches of polish in web UI. Skip it if you need a full design system or brand look rather than interaction details.
-
Get Compound EngineeringCompound Engineering
For developers who want Claude to plan, build, review and ship work in fixed stages with saved plans. Skip it if you prefer quick ad hoc edits without a process.
-
Get security-and-hardening
security-and-hardening
For web developers handling input, logins or personal data who want clear always, ask-first and never rules. Skip it if you need a formal audit report.
-
Get diff-intakediff-intake
For teams adding analytics tracking who want a PR turned into a YAML change brief. Skip it if you just want an ordinary code review of the diff.
-
Get clickhouse-best-practicesclickhouse-best-practices
For teams designing or reviewing ClickHouse tables, queries and inserts against the vendor's own rules. Skip it if you don't run ClickHouse.
-
Get security-reviewsecurity-review
For developers who want a security review of a file or diff that reports only confirmed, exploitable flaws. Skip it if you want every possible weakness listed or a live penetration test.
-
Get swiftui-expert-skillswiftui-expert-skill
For SwiftUI developers who want help writing, reviewing or profiling views, including Instruments traces. Skip it if you are not building with SwiftUI or cannot run Xcode tools on a Mac.
-
Get fixing-accessibilityfixing-accessibility
For front-end developers who want Claude to catch and patch accessibility gaps in buttons, forms and dialogs. Skip it if you need a formal WCAG audit report or automated scanner results.
-
Get golang-error-handling
golang-error-handling
For Go developers who want consistent error wrapping, checks and slog logging in new code or reviews. Skip it if you do not write Go or already follow a company error-handling guide.
-
Get gstackgstack
For people running Garry Tan's gstack workflows on Capy cloud machines or Claude Code. Skip it if you want one standalone skill or do not want an installer that clones and updates code.
-
Get security-reviewsecurity-review
For web developers adding login, payments, uploads or new API routes who want a security checklist at hand. Skip it if you need an automated scan of your code or work outside the Node web stack.
-
Get baseline-ui
baseline-ui
For developers using Tailwind and React who want AI-built screens tidied to a strict house style. Skip it if your project avoids Tailwind or you want bold, decorative visuals.
-
Get better-ui
better-ui
For designers and front-end developers who want exact values for radius, shadows, icons and motion. Skip it if your interface still has broken interactions to fix first.
-
Get better-typography
better-typography
For front-end developers who want text rendering checked and fixed against exact rules. Skip it if you want a new typeface picked for you, which it does only on request.
-
Get Elephant Goldfish ModelElephant Goldfish Model
For engineers who write design docs before building and want fresh-eyed agent reviewers. Skip it if you go straight from idea to code without a written spec.
-
Get Understand AnythingUnderstand Anything
For developers joining or auditing a large codebase who want an explorable map of it. Skip it if your project is small or you can't spend many tokens on the first scan.
-
Get nestjs-best-practicesnestjs-best-practices
For NestJS developers who want Claude to follow set rules on modules, injection and security. Skip it if your server code is in Express, Django or another framework.
-
Get visual-explainervisual-explainer
For developers who want diffs, plans and systems explained as a visual HTML page. Skip it if plain text answers in the terminal suit you fine.
How to choose
code-review-and-quality asks what a change should do, tests the tests by flipping one condition, then labels findings from Critical to FYI and ends with Approve or Request changes. ponytail-review also takes one change, by default your uncommitted work or last commit, and needs a concrete breaking case for every finding.
ponytail-audit reads a whole repository or a named folder instead of one diff, and changes no code. It maps the project, traces data flows and writes up to 20 findings under Must fix, Should fix and Nice to have. thermo-nuclear-code-quality-review is a harsh structural audit of the current branch.
Pick a language skill when you want house rules. react-doctor runs a CLI that scores React code from 0 to 100 and fixes any drop before a commit. rust-best-practices applies Apollo GraphQL's Rust handbook, golang-code-style holds Go style rules, and swiftui-pro reviews SwiftUI and needs Xcode.
review-animations is narrow: you call it by hand on motion code, it checks ten rules such as no transition: all, and it ends in Block or Approve. review-pr from Warp writes review.json for a CI workflow. ast-grep writes syntax-tree search rules, and simplify cleans up code changed in the session.
Secondary picks review one area or add a process. vercel-react-best-practices covers React performance, security-audit looks for security flaws, Compound Engineering has a ce-code-review stage, and Understand Anything maps a codebase into a graph and can review a diff. At least one scanner rated Sentry's security-review and gstack high risk.