Pick a code-review skill without installing five of them
Agent skills chatter stays busy; code-review skills top many install lists. A 3 October 2026 guide to matching one Skill Rank entry to your review habit — not stacking every reviewer you find.
People keep installing Agent Skills for the same job: review the diff before merge. On 3 October 2026 that demand is easy to see without inventing a fake leaderboard. X recent-post counts for agent/cursor/claude skills language totalled about 3.3k posts over the last week (query checked 2026-10-03 Asia/Riyadh morning). A narrower query tying code review to skill/Cursor/Claude language was about 520 in the same window. Public roundups such as Agensi’s Cursor skills list still put code-review style packages near the top of install talk. That is enough signal for a selection guide — not another padded five-skill dump.
The problem is overlap, not scarcity
Skill Rank already lists several review-shaped skills. Installing all of them at once usually means three descriptions fight for the same turn, and you get duplicate severity lists instead of one clear workflow.
Pick one primary review skill, then optionally one feedback skill if you also answer PR comments. Leave the rest for later.
Four Skill Rank entries that cover most cases
These are already on Skill Rank — no catalog add for today’s post:
| Skill | Repo | Use when |
|---|---|---|
| requesting-code-review | obra/superpowers | You want a structured ask before someone (or another agent) reviews |
| receiving-code-review | obra/superpowers | You need a calm process for reading feedback and deciding what to change |
| code-review-and-quality | addyosmani/agent-skills | You want a five-axis review, severity labels, and change-sizing habits |
| open-code-review / code-reviewer | alibaba/open-code-review, google-gemini/gemini-cli | You want a dedicated reviewer persona or CLI-oriented review skill |
Also nearby on the board: caveman-commit and git-commit help after the review, not instead of it. Keep commit skills slash-only if auto-invocation has been noisy (see yesterday’s disable-model-invocation note).
A one-minute decision tree
- You write the code and need a pre-merge pass → start with
code-review-and-qualityorcode-reviewer/open-code-review. - You mainly triage human or bot comments →
receiving-code-review(and optionally a PR-comments skill later). - Your team lacks a shared review checklist →
requesting-code-reviewso the ask is consistent. - You already have one of the above installed → stop. Read its
SKILL.md, run it on a real PR, then decide if a second skill adds a new axis.
Install hygiene that matches the demand
npx skills add addyosmani/agent-skills --skill code-review-and-quality(Example only — pick one repo/skill after reading its SKILL.md.)
Before you trust install counts on a third-party list:
- Open the skill file; check whether it runs git/gh commands or only advises.
- Prefer skills that state severity and stop conditions over skills that always rewrite the whole PR.
- If the skill can commit, push, or merge, treat it as side-effect work: explicit
/invoke, not auto-on.
Limits of this note
- X counts are recent-search volume, not Google Search Console demand or Skill Rank scores.
- Agensi and similar roundups are editorial rankings, not PocketPlay scores — use them as discovery, then verify on the repo and on Skill Rank.
- Stars and install badges do not prove a skill is safe or that two review skills will cooperate.
Sources & checks
How we write and check articles: Editorial method