Skip to content
Back to skills

Coderabbit Review Skill

ASecurity

Use when the user asks for a CodeRabbit review or mentions the coderabbit CLI. Runs `coderabbit review --agent` on local git changes, validates findings against the repo, and reports; applies fixes only when asked.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 7, 2026
toolsbashgit

Works with

  • cli

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned November 7, 2026

npx -y skills add BumpyClock/agent-templates --skill coderabbit-review-skill --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Coderabbit Review Skill?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Coderabbit Review Skill
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/bumpyclock-coderabbit-review-skill/badge)](https://www.skillsdirectory.com/skills/bumpyclock-coderabbit-review-skill)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: coderabbit-review-skill
description: Use when the user asks for a CodeRabbit review or mentions the coderabbit CLI. Runs `coderabbit review --agent` on local git changes, validates findings against the repo, and reports; applies fixes only when asked.
---

# CodeRabbit Review

CodeRabbit CLI reviews local git changes. It is slow (several minutes): run it in the background or with timeout ≥ 10m and keep working while it runs.

## Commands

```bash
coderabbit review --agent                   # Structured findings for agents (default here)
coderabbit review --base main               # Compare against a base branch
coderabbit review --type uncommitted        # Scope: all | committed | uncommitted
coderabbit review --dir <path>              # Only changes inside a directory
coderabbit review findings                  # Re-show last review WITHOUT re-running
coderabbit review --light                   # Faster, reduced-context pass
```

Auth errors → `coderabbit auth login`. Before re-running a review, check `coderabbit review findings` first — re-runs are expensive.

## Workflow

1. **Run** the review scoped to what the user asked about (default: all local changes).
2. **Triage**: dedupe, then order by severity — `critical` → `major` → `potential_issue` → `refactor_suggestion`. For `nitpick` findings, judge each rather than blanket-skip: worth keeping if it's trivial and touches code already being changed, or the user asked for nitpicks; otherwise skip. Give a one-line reason either way in the report.
3. **Validate**: CodeRabbit produces false positives. Confirm each finding against the actual code (claimed path/symbol/behavior exists, issue is reachable) before treating it as real. Many findings → fan out read-only sub-agents; a handful → verify inline.
4. **Report**: valid findings with location + smallest safe fix, invalid ones with the evidence that disproves them. Scan all findings — including skipped nitpicks — for clusters: several of the same shape usually mean one root cause (missing convention, missing abstraction, architectural smell). Surface the pattern and propose the holistic fix, not per-file patches.
5. **Fix only when the user asked for fixes.** Then follow normal change discipline: valid findings only, tests for behavior changes, verify before claiming done.

Files in this skill

  • SKILL.md2.3 KB
  • agents/openai.yaml294 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.