Codebase Argus is a portable multi-agent review playbook designed to provide evidence-first analysis of GitHub pull requests, CI failures, and complex fork integrations.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install codebase-argus
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 codebase-argus 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 Codebase Argus?
Codebase Argus is a highly adaptable agent skill that brings rigorous, evidence-based oversight to your development workflow. By leveraging Openclaw Skills, it automates the heavy lifting of reviewing GitHub PRs and diagnosing CI failures. It is uniquely designed to act as a tribunal, allowing multiple AI providers—like OpenAI, Claude, and Gemini—to cross-reference their findings for maximum accuracy.
Beyond simple reviews, Codebase Argus excels at managing long-lived forks. It can simulate merges and rebases to identify risks before they hit your main branch. Whether deployed as a GitHub App webhook or used as a local CLI tool, it prioritizes concrete evidence over model hallucinations, ensuring that security-critical changes and breaking workflows never go unnoticed.
Codebase Argus Use Cases
- Reviewing high-stakes GitHub Pull Requests for security flaws, missing tests, or sensitive workflow edits.
- Diagnosing root causes of CI failures by analyzing GitHub Actions logs and local failure output.
- Generating safe autofix plans for mechanical failures like lockfile refreshes or linting errors.
- Managing long-lived forks by simulating upstream syncs and identifying merge/rebase conflicts.
- Deploying a GitHub App to provide automated /argus command responses on repository comments.
How Codebase Argus Works
- Evidence Collection: The agent fetches PR metadata, diffs, GitHub Action logs, or local CI output to build a comprehensive context package.
- Policy Evaluation: It checks the code against local .codebase-argus.yml policy files to ensure compliance with repository-specific rules.
- Tribunal Analysis: The evidence is sent to one or more AI providers (OpenAI, Anthropic, Gemini, or local CLIs) for a multi-agent review.
- Risk Assessment: The skill identifies high-risk areas such as auth changes, route-handling edits, or dependency updates.
- Reporting & Feedback: Findings are presented as a Markdown report, a CLI summary, or posted as inline comments via a GitHub Webhook.
Codebase Argus Setup
Ensure you have node, npm, and git installed. Clone the repository and install dependencies:
npm install
# Run a basic review on a public PR
npm run argus -- review owner/repo#123
To enable AI-powered reviews, export your provider keys:
export OPENAI_API_KEY='your-key'
export ANTHROPIC_API_KEY='your-key'
# Run a multi-agent tribunal
npm run argus -- review owner/repo#123 --tribunal openai-api,anthropic-api
Codebase Argus Data Schema & Taxonomy
Codebase Argus organizes data within a local cache and utilizes environment variables for secure authentication.
| Data Type | Description |
|---|---|
| Local Cache | Located in .cache/repos, used for temporary worktrees and log analysis. |
| Policy File | .codebase-argus.yml defines the local rules and gates for the agent. |
| Environment | Uses GITHUB_TOKEN for API access and provider-specific keys for LLM analysis. |
| Webhook State | Managed via GitHub Labels (e.g., argus:paused) to control automation flow. |
Codebase Argus Advanced Features
- Tribunal Mode: Aggregates insights from multiple LLM providers to reduce false positives and verify facts.
- Downstream Sync Simulation: Uses
git merge-treeandgit range-diffto evaluate integration risks without altering the branch. - GitHub App Integration: Supports slash commands (
/argus review,/argus autofix) directly within PR comments. - Autofix Planning: Generates narrow, mechanical repair plans for automated cleanup tasks.
- Local Dashboard: Provides a visual interface via
npm run devto inspect merge risks and agent workflows.
SKILL.md
Loading
Related Openclaw Skills
Home Emergency Shutoff Map
A specialized AI workflow for documenting critical home utility shutoffs, safety equipment, and emergency contacts into a printable orientation map.
wc-tool
A robust multi-file utility for counting lines, words, characters, and bytes with structured output support.
Video Downloader
A powerful command-line utility for downloading high-resolution video and audio content from various web platforms.
wc-chars-tool
A robust Unicode-aware character counting utility that handles CJK characters, grapheme clusters, and detailed text metrics.
Credit Card Chargeback Prep Kit
A comprehensive workflow to organize factual evidence and documentation for credit card transaction disputes and chargebacks.
AI Agent Permission Planner
A specialized security framework designed to build least-privilege permission plans before connecting AI assistants to sensitive tools and data.
Featured*
