agentleFS
BSSE23047BSSE23047/nauraiz-dev/skills/review/SKILL.md

review

Skill2 starsChanged 15 days agoMITView on GitHub

Review code before or in a PR: 'review my branch', 'check this before the PR', 'review PR #12'. Runs the reviewer agent on the diff and the gate; writes review notes, never an approval. Not for pushing or opening a PR (that is /nauraiz-dev:ship) or fixing findings.

192 other projects use a version of this.

Adopted by

Most recently taken up by

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.
  • dzhngSkill

    close-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

  • dzhngSkill

    design-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

  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新

    Also in 5 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 5 projects that use this

  • +3 more from dzhng
  • +5 more from Lambenthan

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 1 is distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 157 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Never paste a diff into this thread.

    line 18
  2. Very specific:

    Never change the owner's checkout: no git checkout, git switch, gh pr checkout, git stash, no edits.

    line 22
  3. Very specific:

    Run git diff origin/<base>...HEAD > "<diff>".

    line 37
  4. Distinctive:

    Do not fix anything unless the owner asks.

    line 46
  5. Very specific:

    When isCrossRepository is true, before anything runs, warn: this runs the fork's code through the gate.

    line 63
  6. Very specific:

    If the gate needs dependencies, install them in <wt> (npm ci, flutter pub get).

    line 65
  7. Very specific:

    If the /code-review skill is available, run it on PR <n> (no --comment).

    line 71
  8. Very specific:

    Only after the owner says "merge": gh pr merge <n> --<branch.merge> (merge, squash or rebase), never --admin.

    line 75

“Common” is measured: kept by at least 6 in 10 of the 157 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

review

Two modes. No argument, or the owner's own branch: branch mode. A PR number: PR mode.

Config: .claude/nauraiz-dev.json, with .claude/nauraiz-dev.local.json over it. Keys: branch.base (default main), branch.merge (default merge), models.review, gate.

Rules for both modes:

  • "Stop" means: end your turn and wait.
  • Never paste a diff into this thread. Write it to a file under git rev-parse --git-path and give the path to the reviewer; look at --stat only.
  • These are review notes. Never write "Approved" or "LGTM", never submit an approving GitHub review.
  • Never change the owner's checkout: no git checkout, git switch, gh pr checkout, git stash, no edits. Note git branch --show-current first; it must match at the end.
  • Merge only when the owner says "merge" in this session.

Reviewer: the Agent tool with subagent_type nauraiz-dev:reviewer. Pass model only when models.review is sonnet, opus, haiku or fable; otherwise leave it out (the agent's own model). Dispatch only after the diff file is written. Its prompt: the diff file path, the base, the plan path and issue if any, and for a PR its title and body. It has no shell and works from the diff file; it returns findings by severity.

Branch mode

  1. git fetch origin <base>. If git status --short lists changes, say they are not in the diff review, and that the gate result includes uncommitted changes.
  2. Diff file: <diff> is what git rev-parse --git-path nauraiz-dev-review.diff prints. Run git diff origin/<base>...HEAD > "<diff>". Empty: say "Nothing to review against origin/<base>" and stop.
  3. Dispatch the reviewer on <diff>. Meanwhile run node "${CLAUDE_PLUGIN_ROOT}/scripts/gate/run.mjs" once, from any folder in the project (it prints every exit code: never append echo $?, pipes or cd; re-run only after you changed code, or once with --plan <path> when a row says no plan for this branch). Exit 2: as the gate skill says.
  4. Report: the findings ranked by severity, the gate output verbatim, what was not checked. End with what to fix before /nauraiz-dev:ship, or "No blocking findings". Do not fix anything unless the owner asks. Stop.

PR mode

  1. gh pr view <n> --json number,title,body,author,baseRefName,headRefName,isCrossRepository,closingIssuesReferences.
  2. <diff> from git rev-parse --git-path nauraiz-dev-pr-<n>.diff; run gh pr diff <n> > "<diff>". Dispatch the reviewer on it.
  3. Gate on the PR head, in a temporary worktree outside the repo:
    • <tmp> is what node -e "console.log(require('os').tmpdir())" prints. <wt> is <tmp>/<repo>-review-<n>-<random> (<repo> this folder's name, <random> six hex characters you pick), never relative to the current folder. Quote it.
    • <wt> must not exist (git worktree list and the file system); if it does, stop.
    • git fetch origin pull/<n>/head, then git worktree add --detach "<wt>" FETCH_HEAD.
    • In <wt>, delete <wt>/.claude/nauraiz-dev.json and <wt>/.claude/nauraiz-dev.local.json, then copy this checkout's two files (those that exist) into <wt>/.claude/. The gate commands are then the owner's, but the PR still controls what they run (package.json scripts, tests, build files).
    • When isCrossRepository is true, before anything runs, warn: this runs the fork's code through the gate; ask, and stop until the owner answers.
    • If the gate needs dependencies, install them in <wt> (npm ci, flutter pub get).
    • node "${CLAUDE_PLUGIN_ROOT}/scripts/gate/run.mjs" --cwd "<wt>" (once; no echo $?, pipes or cd); add --plan "<wt>/<plan>" if the PR carries a plan for headRefName.
    • Then, only if git worktree add succeeded, git worktree remove --force "<wt>" with the exact <wt> this run created, never another path, even when a step failed.
  4. If the /code-review skill is available, run it on PR <n> (no --comment).
  5. Review notes: the reviewer's and /code-review's findings merged and ranked by severity, the gate output verbatim, what was not verified. Post them on the PR (gh pr comment) only if the owner asks. Stop.
  6. Only after the owner says "merge": gh pr merge <n> --<branch.merge> (merge, squash or rebase), never --admin. After a merge only: close any of closingIssuesReferences still open (gh issue close), and offer the next kickoff prompt with the kickoff-prompt skill.

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.