An adversarial AI code reviewer designed to ruthlessly identify security flaws, technical debt, and bad practices across multiple programming languages.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install critical-code-reviewer
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 critical-code-reviewer 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 Critical Code Reviewer?
The Critical Code Reviewer is a specialized persona for AI agents that functions as a senior engineer with zero tolerance for mediocrity. Unlike standard linters, this skill adopts a skeptical mindset, assuming every line of code is broken or inefficient until proven otherwise. It is designed to protect codebases from entropy by identifying "slop," lazy naming conventions, and structural failures that human reviewers might overlook.
By leveraging Openclaw Skills, developers can automate the process of rigorous peer review. The skill specializes in Python, R, JavaScript, TypeScript, SQL, and CSS, ensuring that error handling, type safety, and performance are prioritized. It evaluates the actual artifact rather than the developer's intent, providing a constructive but brutal assessment that enforces high engineering standards.
Critical Code Reviewer Use Cases
- Performing a rigorous pre-PR audit to catch edge cases and race conditions before they reach production.
- Identifying security vulnerabilities such as SQL injection, XSS, and unhandled promise rejections.
- Refactoring "junk drawer" files and cleaning up disorganized component structures in frontend frameworks.
- Validating type safety in TypeScript and ensuring proper error handling in Python or R scripts.
- Scrubbing "cargo cult" patterns and redundant comments from legacy codebases.
How Critical Code Reviewer Works
- The agent analyzes the provided code snippet or pull request using a "guilty until proven exceptional" framework.
- It scans for specific "Slop" patterns, including lazy naming (e.g.,
data,temp) and redundant comments that insult the reader's intelligence. - The skill applies an adversarial lens to detect hidden risks like unhandled nulls, malformed API responses, and silent failures.
- Issues are categorized into four distinct severity tiers: Blocking, Required Changes, Strong Suggestions, and Noted style points.
- The agent generates a structured report containing a BLUF (Bottom Line Up Front) summary, file-specific line references, and a final verdict (Approve or Request Changes).
Critical Code Reviewer Setup
To integrate this capability into your workflow using Openclaw Skills, you should include the skill definition in your agent's system prompt or skill directory.
# Example of adding the skill to your local environment
mkdir -p .ai/skills
curl -o .ai/skills/critical-code-reviewer.md https://raw.githubusercontent.com/posit-dev/skills/main/posit-dev/critical-code-reviewer/SKILL.md
Once added, you can trigger the review by asking the agent to "critique my code" or "find issues in this PR."
Critical Code Reviewer Data Schema & Taxonomy
The skill produces a structured Markdown response with the following sections:
| Section | Description |
|---|---|
| Summary | A high-level assessment of the code's quality and risk profile. |
| Critical Issues | Blocking errors such as security holes and logic failures. |
| Required Changes | Necessary fixes for technical debt, lazy patterns, and slop. |
| Suggestions | Recommendations for performance or architectural improvements. |
| Verdict | Final status: Request Changes, Needs Discussion, or Approve. |
| Next Steps | Actionable items for the user, such as discussing issues or adding comments to a PR. |
Critical Code Reviewer Advanced Features
- Multi-language support covering Python, R, JavaScript, TypeScript, and SQL.
- Context-aware partial reviews that flag "Verify" items when full codebase context is missing.
- Integration with the AskUserQuestion tool for interactive troubleshooting of identified issues.
- Automated PR commenting capabilities with proper attribution to the skill.
- Adversarial detection patterns specifically tuned for React, Vue, and Svelte component architecture.
SKILL.md
Loading
Related Openclaw Skills
MyVibe Publish
A comprehensive automation tool for publishing static web content, directories, and ZIP archives directly to the MyVibe platform.
Web Scraper Pro (zhuyu28)
An advanced browser automation skill for complex data extraction, form filling, and UI interaction.
NeuralMemory
A biologically-inspired associative memory system that gives AI agents persistent, intelligent recall using spreading activation graphs.
Multi-Agent Coordinator
A robust orchestration layer for managing specialized AI agents to complete complex, multi-step tasks efficiently.
Describe Design
A powerful AI agent skill to automatically research codebases and generate structured architectural documentation with Mermaid diagrams.
LinkedIn Automator
An automated LinkedIn growth tool for content creation, scheduling, and performance tracking using browser-based workflows.
Featured*
