agentleFS
k3ybladewielderk3ybladewielder/skills/pr-reviewer/SKILL.md

pr-reviewer

Skill7 starsChanged 53 days agoApache-2.0View on GitHub

Automatically reviews PRs -> checks out the branch, analyzes commits and diffs, generates a structured report, and asks if it can comment on the PR.

22 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 22 edited versions we read keep these.

  • 7

    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:

    When user ask for a PR review you must follow the guidelines bellow.

    line 9
  2. Distinctive:

    If the user does not explicitly provide the repository, branch name, and PR number, you MUST request them before performing any action.

    line 14
  3. Very specific:

    Fetch PR commits: Use gh pr view <number> --json commits,title,body,headRefName,baseRefName to get all commits

    line 19
  4. Distinctive:

    If there are .parquet or binary files: Use Python to inspect schema/columns

    line 21
  5. Distinctive:

    ASK the user: if the report is good or needs changes before commenting

    line 23
  6. Distinctive:

    Write the report in English, unless requested in Portuguese

    line 100
  7. Distinctive:

    Don't block for style — focus on bugs, security, and architecture

    line 103
  8. Distinctive:

    If the PR is large (>500 lines), organize by module/concern

    line 104

“Common” is measured: kept by at least 6 in 10 of the 22 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

PR REVIEW SKILL

When user ask for a PR review you must follow the guidelines bellow. USER must provide:

  • The repository (subproject within the workspace)
  • The branch name of the PR
  • The PR number (e.g., #35)

If the user does not explicitly provide the repository, branch name, and PR number, you MUST request them before performing any action.

Mandatory process:

  1. Fetch and checkout: git fetch origin <branch> and git checkout <branch> in the indicated repository
  2. Fetch PR commits: Use gh pr view <number> --json commits,title,body,headRefName,baseRefName to get all commits
  3. Analyze each commit: For every commit, run git show <sha> --format="" --stat and git show <sha> --format="" to view the full diff
  4. If there are .parquet or binary files: Use Python to inspect schema/columns
  5. Generate the report in the format below
  6. ASK the user if the report is good or needs changes before commenting
  7. Only after confirmation: Post as a comment on the PR via gh pr comment <number> --body '<report>'

Report format:

## 🔍 Code Review — PR #<number>: <title>

### Priority Scale

| Level | Meaning | Expected Action |
|-------|---------|----------------|
| **P1** | Blocker — bug, security issue, or breaking functionality | Must be fixed before merge |
| **P2** | Important — real issue but not critical | Strongly recommended to fix |
| **P3** | Improvement — design smell, technical debt, desirable refactor | Could be a follow-up |
| **P4** | Cosmetic — style, DRY, optional suggestions | Not blocking, up to the author |

---

### Commit Summary (<N> commits)

| # | SHA | Message | Impact |
|---|-----|---------|--------|
| 1 | `<sha7>` | <message> | <short description of impact> |
| ... | ... | ... | ... |

**Base:** `<base branch>` (PR #<base PR number, if any>)

---

### Files Changed

| File | Type | Change |
|------|------|--------|
| `<path>` | New/Modified/Removed/Renamed | <short description> |

---

### ✅ Positive Points

1. **<title>**: <explanation>
2. ...

---

### ⚠️ Points of Attention

#### P1 — <short title>
<detailed explanation of the issue, why it matters, and solution suggestion>

#### P2 — ...

---

### 📋 Verdict

<general summary in 2–3 lines>

- **P1 (priority)**: <summary>
- **P2 (priority)**: <summary>

**<Approvable/Approvable with notes/Requires changes>. <Main point>.**

Review Criteria:

  • Security: Hardcoded credentials, exposed secrets, eval with user input
  • Quality: Duplicate code, inline vs top-level imports, private methods used externally
  • Performance: Duplicate file reads, unnecessary loops, temporary objects not cleaned up
  • Maintainability: Hardcoded logic that doesn't scale, tight coupling
  • Completeness: Features created but not integrated, dead code
  • Documentation: Outdated docs, missing docstrings for public APIs
  • Tests: Adequate coverage, reusable fixtures, correct markers
  • CI/CD: Robust workflows, error handling, concurrency

Rules:

  • Write the report in English, unless requested in Portuguese
  • Be constructive — point out problems AND suggest solutions
  • Prioritize issues: P1 (blocker), P2 (important), P3 (improvement), P4 (cosmetic)
  • Don't block for style — focus on bugs, security, and architecture
  • If the PR is large (>500 lines), organize by module/concern
  • ALWAYS ask the user before posting the comment

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.