Best Claude Skills for Code Review in 2026: 15 Compared
Top 15 Claude skills for code review. The #1 and #2 entries are a paired workflow (receiving + requesting review) from obra/superpowers — install both together.
Code Review as a category reveals an underappreciated truth: the act has two roles, not one. The top two skills in this list are literally that pair — receiving-code-review (for the author) and requesting-code-review (for the reviewer) — both from obra/superpowers, both at 167K signal and 4/5 quality. Install them together or you only get half the workflow. The rest of the top 15 is a mix of legitimate review tools (Matt Pocock's Grill Me, wpank's systematic patterns) and some skills that got tagged code-review but really belong elsewhere (#3–#5 are agent-delegation skills, not review tools).
Quick Pick
Install receiving-code-review + requesting-code-review together (obra/superpowers). The first is for when you're acting on review feedback; the second is for when you're producing a PR that needs review. They're complementary halves of the same workflow.
What These Skills Actually Do
Code review skills cluster into four patterns: (1) Paired workflow skills (#1, #2) — methodology, not tooling. Encode how to receive feedback (without performative agreement) and how to request it (before merge, against requirements). Both at 4/5 quality. (2) Socratic review (#6 Grill Me, #7 Matt Pocock's bundle) — instead of telling you what's wrong, asks hard questions about your design. Most effective for surfacing hidden complexity. (3) Automated PR review (#9 wpank's Code Review, #10 pr-reviewer, #11 code-review-fix, #15 LogicArt) — diff analysis, lint integration, security checks. Cheaper than a human reviewer; less judgmental than one too. (4) Misclassified entries (#3–#5) — Hermes's delegate-to-claude-code, codex, opencode skills got tagged with code-review but they're really about delegating coding to other agents. Treat them with skepticism in this context. What separates great review skills from mediocre ones: do they make the reviewer think more (Grill Me does), or just produce more text (most auto-review skills do)?
How We Ranked
We sorted 15 candidate skills by a composite score:
- Popularity signal — the highest of GitHub stars, install count, or ClawHub download count. Log-scaled so a 100-star skill doesn't get buried under a 100,000-star one if the smaller one is meaningfully better.
- Quality score — when set, a 0–5 rubric that breaks ties within popularity tiers. Roughly 15% of catalog skills carry a quality score today; we surface it in the comparison table when available.
The formula is identical across the entire Best-Of 2026 series, so you can compare apples to apples between categories.
The Top 15
1. receiving-code-review
Skill · obra/superpowers · 167.5K signal · quality 4/5 For acting on review feedback — technical rigor over performative agreement.
The take: Designed for the author of code under review. The killer feature: explicit "before implementing suggestions, especially if feedback seems unclear or technically questionable" framing forces Claude to verify rather than just agree. Prevents the failure mode where every review comment gets dutifully implemented and the final code is worse than the first draft.
2. requesting-code-review
Skill · obra/superpowers · 167.5K signal · quality 4/5 For requesting review before merge — verify work meets requirements.
The take: The reviewer side of the pair with #1. Forces a structured pre-merge check against requirements before asking a human (or another agent) to review. Saves reviewer-round-trip time. Install with #1; they're complementary halves.
3. claude-code (Hermes)
Skill · NousResearch/hermes-agent · 124.8K signal · quality 4/5 Delegate coding to Claude Code CLI.
The take: Tagged code-review but actually about delegating coding to another Claude instance — not a review skill at all. The high signal reflects how popular Hermes's repo is overall. Treat as not really belonging on this list; install it for orchestration scenarios, not for review.
4. codex (Hermes)
Skill · NousResearch/hermes-agent · 124.8K signal · quality 4/5 Delegate coding to OpenAI Codex CLI.
The take: Same caveat as #3. Useful if you want Claude to spawn Codex subtasks; not what you install for code review.
5. opencode (Hermes)
Skill · NousResearch/hermes-agent · 124.8K signal · quality 4/5 Delegate coding to OpenCode CLI.
The take: Same again. The "PR review" mention in the description is what got it tagged into this list; the actual implementation is agent delegation.
6. Grill Me (Code Review)
Command · mattpocock/skills · 49.0K signal · quality 4/5 Socratic code review by Matt Pocock — asks hard questions to surface hidden complexity.
The take: The single best alternative to the Superpowers pair (#1/#2) for actual review work. Socratic approach means it doesn't tell you what's wrong — it asks questions you can't easily dodge. Pocock's TypeScript credibility shows in the depth.
7. Matt Pocock TypeScript Skills
Plugin · mattpocock/skills · 43.5K signal · quality 4/5 TypeScript-specific skills bundle — code review, TDD, architecture improvement, advanced TS patterns.
The take: Broader bundle that includes Grill Me. Install this instead of #6 if you also want TDD and architecture skills from the same author. TypeScript-team essential.
8. Agent Skills by Addy Osmani
Plugin · addyosmani/agent-skills · 21.0K signal · quality unrated 20 production-grade engineering skills — define, plan, build, verify, review, ship.
The take: The "review" stage of Osmani's lifecycle bundle. Slightly less focused than the Pocock skills (#6, #7); broader scope if you want one bundle to cover everything.
9. Code Review (wpank)
Skill · wpank/code-review · 11.8K signal · quality unrated Systematic patterns — security, performance, maintainability, correctness, testing. Severity levels, structured feedback, anti-patterns.
The take: The most structured automated review skill — severity levels are the differentiator over the lighter alternatives at #10/#11. Install when you want consistent rubric-based review output you can audit later.
10. pr-reviewer
Skill · briancolinger/pr-reviewer · 6.4K signal · quality unrated GitHub PR review — diff analysis, lint integration, structured reports, security checks.
The take: PR-flavored automation; install if you want Claude to auto-comment on PRs in CI. Doesn't think hard — produces clean reports.
11. code-review-fix
Skill · landyun/code-review-fix · 2.7K signal · quality unrated Auto-review + fix suggestions + optionally apply repairs.
The take: Goes one step further than #10 — also offers fixes. Useful for the "obvious" findings (formatting, simple bugs); dangerous for the non-obvious ones. Keep the "apply" feature off until you trust it.
12. Security Audit (Sona)
Skill · virtaava/sona-security-audit · 2.6K signal · quality unrated Fail-closed security auditing — trufflehog secrets, semgrep SAST, prompt-injection signals, supply-chain hygiene.
The take: Specifically for auditing the skills you install, not for general code review. Tagged code-review because it does review code, but the angle is supply-chain security. Worth installing if you regularly add unverified skills.
13. Compile LaTeX & Typst (TypeTex)
Skill · gregm711/typetex · 2.5K signal · quality unrated Compile Typst/LaTeX to PDF via API.
The take: Doesn't belong in code-review — got tagged via the "compile" verb. Skip in this context; it's a PDF rendering skill.
14. Clauditor
Skill · apollostreetcompany/clauditor · 2.4K signal · quality unrated Tamper-resistant audit watchdog with HMAC-chained evidence logs.
The take: Forensic-grade activity logging for agents — not code review in the human sense. Install when running unattended agents you need verifiable audit trails for.
15. Code Review (LogicArt)
Skill · jpaulgrayson/quack-code-review · 2.4K signal · quality unrated Code analysis via LogicArt — bugs, security, logic-flow visualizations.
The take: The visualization angle (logic flow graphs) is the differentiator. Useful when reviewing unfamiliar codebases where you want a structural overview before diving in.
Comparison Table
| # | Skill | Type | Stars / Installs | Quality | License |
|---|---|---|---|---|---|
| 1 | receiving-code-review | Skill | 167.5K | 4/5 | — |
| 2 | requesting-code-review | Skill | 167.5K | 4/5 | — |
| 3 | claude-code | Skill | 124.8K | 4/5 | MIT |
| 4 | codex | Skill | 124.8K | 4/5 | MIT |
| 5 | opencode | Skill | 124.8K | 4/5 | MIT |
| 6 | Grill Me (Code Review) | Command | 49.0K | 4/5 | MIT |
| 7 | Matt Pocock TypeScript Skills | Plugin | 43.5K | 4/5 | MIT |
| 8 | Agent Skills by Addy Osmani | Plugin | 21.0K | — | MIT |
| 9 | Code Review | Skill | 11.8K | — | — |
| 10 | pr-reviewer | Skill | 6.4K | — | — |
| 11 | code-review-fix | Skill | 2.7K | — | — |
| 12 | Security Audit (Sona) | Skill | 2.6K | — | — |
| 13 | Compile LaTex & Typst into PDF with TypeTex | Skill | 2.5K | — | — |
| 14 | Clauditor | Skill | 2.4K | — | — |
| 15 | Code Review | Skill | 2.4K | — | — |
FAQ
How is this list different from the category page on aiskill.market?
The category page is a directory: every skill in the category, sortable and filterable. This list is editorial — opinionated, time-stamped (2026-05-18), and ranked. Use the directory when you know what you want; use this when you don't.
Why does the #1 pick have fewer stars than #5?
Stars are one signal among several. The composite score above also includes install counts (which reflect actual usage on aiskill.market) and the optional quality score. A skill with a smaller star count can rank higher if its installs or quality score are strong enough to offset.
Are these all free?
Open source. Sona Security Audit (#12) requires trufflehog and semgrep installed locally (free OSS). LogicArt's #15 may have backend dependencies — check the upstream README.
How do I install one?
Each linked skill page has install instructions. The fastest path is the one-line install via the aiskill.market CLI or by adding the source repo as a Claude Code plugin marketplace.
How often does this list update?
Quarterly. Code-review tooling moves slowly; expect ranking shifts within the top 15 but not new entrants displacing #1-#2.
Why are #3-#5 on this list if they're not really review skills?
Tag misclassification. The Hermes repo organizes its skills with broad keyword tagging, which earns those entries the code-review tag even though their actual function is coding delegation. The ranking is honest about install signal; the editorial pass flags the mismatch.
What's the right starter pack for code review?
Three skills: #1 receiving-code-review + #2 requesting-code-review (workflow pair) + #6 Grill Me (Socratic depth). Add #9 wpank's Code Review if you want a fourth that adds structured rubric-based output.
Related Categories
- Best Claude Skills for Testing & TDD in 2026 — Superpowers' TDD skill comes from the same author
- Best AI Skills for Development & Code Tools in 2026 — broader development context
- Best AI Skills for Security & Privacy in 2026 — Sona Security Audit (#12) overlaps
Browse The Full Catalog
Find every skill in this category — including the ones that didn't make the top 15 — at the main browse page.
Part of the Best-Of 2026 series. Updated 2026-05-18. Skills sampled from a catalog of ~262 active entries with a combined 853.7K popularity signal across the ranked entries.