agentleFS
IlyaSemenovIlyaSemenov/agent-review-skill/skills/agent-review/SKILL.md

agent-review

Skill4 starsChanged 15 days agoMITView on GitHub

Run a peer CLI agent (Claude Code, Codex, OpenCode, pi), or prepare a reviewer handoff via user relay, to review code, diffs, or plans, including a plan through its implementation and commits, when explicitly requested by the user or via $agent-review.

11 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use commit.

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.
  • SentrySkill

    commit

    Create commit messages following Sentry conventions. Use when committing code changes, writing commit messages, or formatting git history. Follows conventional commits with Sentry-specific issue references.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

10 rules. 0 are common, 9 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 9

    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
  1. Distinctive:

    Use a peer agent as a reviewer, not as the authority.

    line 8
  2. Distinctive:

    Keep responsibility for deciding which feedback to accept, reject, or discuss.

    line 9
  3. Distinctive:

    If the user explicitly requests user relay, read Review via user relay and Review loop.

    line 16
  4. Distinctive:

    If neither user relay nor a CLI reviewer is named, or the choice is ambiguous, ask the user instead of guessing.

    line 20
  5. Distinctive:

    When running the CLI path, also read the document matching the host environment:

    line 37
  6. Distinctive:

    If no host document exists, run the helper in the foreground with the required permissions.

    line 42
  7. Distinctive:

    If multiple targets are plausible or the request is underspecified, ask what to review.

    line 48
  8. Distinctive:

    If the subject is a plan that you will implement yourself, also follow Plan and implementation.

    line 51
  9. Very specific:

    Use path-based input for project files that the reviewer can access.

    line 56
  10. Distinctive:

    If the user asks for feedback on this skill or on how a review went, rather than on the reviewed artifact, also read Meta-review.

    line 61

“Common” is measured: kept by at least 6 in 10 of the 9 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Agent Review

Use a peer agent as a reviewer, not as the authority. Keep responsibility for deciding which feedback to accept, reject, or discuss.

Choose the runtime branch

Act as the host agent running the skill. The external reviewer follows the generated review prompt and does not read this skill.

  • If the user explicitly requests user relay, read Review via user relay and Review loop.
  • Otherwise, read Review via CLI, Review loop, and any matching host and reviewer guidance selected below.

Load only the common documents for the selected review path and its applicable host and reviewer documents. If neither user relay nor a CLI reviewer is named, or the choice is ambiguous, ask the user instead of guessing.

CLI reviewers

Available review agents, by their --agent identifier:

  • claude (Claude Code).
  • codex (Codex).
  • opencode (OpenCode) — read OpenCode reviewer guidance.
  • pi (pi) — read pi reviewer guidance.

Map a naturally named reviewer to the identifier on the left. Pass the selected identifier, or manual, through the required --agent flag on every round. CLI review requires Node.js 22.18 or newer, npm, first-run network access, the selected reviewer's CLI on PATH, and its authentication in the normal local environment.

Select host guidance

When running the CLI path, also read the document matching the host environment:

  • Claude Code host: read Claude Code host guidance.
  • Codex host: read Codex host guidance.

If no host document exists, run the helper in the foreground with the required permissions.

Identify the review subject

Review only a clearly identified subject from context, such as a diff, plan, design note, issue summary, code snippet, or project files. In Plan Mode, use the current plan text as the subject. If multiple targets are plausible or the request is underspecified, ask what to review.

State the subject and compose the review focus according to Review focus. If the subject is a plan that you will implement yourself, also follow Plan and implementation.

For repository changes reviewed via CLI, instruct the reviewer to run the exact diff command rather than pasting a captured diff. For a working-tree review, also instruct the reviewer to run git status --short and inspect every in-scope untracked file.

Use path-based input for project files that the reviewer can access. Materialize larger non-repository subjects in /tmp only when the selected review path can access that file.

Meta-review

If the user asks for feedback on this skill or on how a review went, rather than on the reviewed artifact, also read Meta-review.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.