A multi-agent code review system that performs structured analysis on Pull Requests and diffs with high-confidence filtering.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install openclaw-code-review-skill
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install openclaw-code-review-skill using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Code Review?
The Code Review skill is a high-performance automation tool designed for Openclaw Skills that replicates advanced developer workflows. It utilizes a multi-agent architecture to perform parallel audits of code changes, ensuring that feedback is both comprehensive and technically accurate. By leveraging distinct AI personas for logic, compliance, and history, this skill provides a rigorous layer of quality control for any repository.
At its core, the skill focuses on reducing noise by implementing a confidence-based scoring system. Only issues with a confidence rating of 80 or higher are reported, effectively filtering out common AI hallucinations and minor nitpicks. This makes it an essential tool for teams looking to integrate Openclaw Skills into their CI/CD or local development environments to maintain high standards of code integrity.
Code Review Use Cases
- Reviewing Pull Requests before merging to main branches.
- Analyzing local git diffs to catch logic errors before committing.
- Enforcing project-specific coding standards defined in CLAUDE.md files.
- Auditing security vulnerabilities and hardcoded credentials in new code changes.
How Code Review Works
- Target Identification: The skill identifies the review target via a PR URL, local search using the GitHub CLI, or raw diff content.
- Context Collection: It gathers PR metadata, diff files, and repository-specific instructions (CLAUDE.md) to establish a baseline for the review.
- Parallel Agent Spawning: The system spawns three specialized agents: a Compliance Agent (style and rules), a Logic Agent (bugs and security), and a History Agent (Git context and regression).
- Confidence Validation: A secondary validation step verifies findings from the Logic Agent to ensure a confidence score of 80+.
- Structured Reporting: Results are aggregated into a formatted report categorized by risk level (High/Medium) and verified status.
Code Review Setup
To use this skill within Openclaw Skills, ensure you have the GitHub CLI installed and authenticated.
# Install GitHub CLI
brew install gh
# Authenticate with your repository
gh auth login
# Trigger a review in your Openclaw environment
/code-review <PR_URL>
Code Review Data Schema & Taxonomy
| Component | Description | Data Source |
|---|---|---|
| PR Metadata | Title, Author, Files Changed, Lines Added/Deleted | GitHub API via gh CLI |
| Context Rules | Project-specific style guides and constraints | CLAUDE.md |
| Issue Schema | Confidence Score, File Path, Line Numbers, Risk Category | Multi-Agent Output |
| History Data | Git blame and log context for affected files | Local Git CLI |
Code Review Advanced Features
- Multi-Model Orchestration: Utilizes different models (e.g., Opus for complex logic, Sonnet for style) to optimize for speed and accuracy within Openclaw Skills.
- Automated GitHub Integration: Ability to post the final review directly as a comment on the GitHub Pull Request.
- CLAUDE.md Adherence: Deep integration with project-specific markdown rules to ensure brand or architecture consistency.
- Noise Filtering: Built-in logic to ignore issues that existed prior to the current PR or changes that are handled by standard Linters.
SKILL.md
Loading
Related Openclaw Skills
Contradiction Detector
A diagnostic tool designed to identify and resolve conflicting instructions across multiple agent files to prevent AI hallucinations and inconsistent behavior.
Zhang Zara Perspective
A strategic thinking lens for navigating AI-era careers, product development, and personal leverage through a builder-first methodology.
Journal Monitor
An automated research assistant skill that monitors academic journals via ISSN and generates instant bilingual English-Chinese reports.
Pilot Event Replay
A specialized utility for capturing event streams to NDJSON files and replaying them to simulate workloads or debug downstream consumers.
sf-ai-agentforce-testing
A professional testing framework for Agentforce agents featuring dual-track validation, multi-turn conversation checks, and automated coverage analysis.
flight-scout
A professional-grade aviation data skill for executing complex flight searches, price monitoring, and hidden-city ticketing queries via commercial APIs.
Featured*
