Self Review

What This Skill Does

The Self-Review skill is a lightweight, rule-based quality assurance agent designed for OpenClaw. Developed by leic8959-sudo, this utility acts as a local gatekeeper for your agent's outputs. Before any response is delivered to the end-user, the skill evaluates the content against four fundamental pillars: clarity, conciseness, actionability, and structure. Unlike resource-heavy alternatives that rely on remote LLM endpoints, this skill operates as a deterministic engine, meaning it processes text locally without incurring additional API costs or latency overhead. It returns a simple exit code (0 for approved, 1 for needs improvement), making it an ideal candidate for automated CI/CD-style pipelines for agent development.

Installation

To integrate this skill into your environment, utilize the OpenClaw CLI tool. Run the following command in your terminal: clawhub install openclaw/skills/skills/leic8959-sudo/self-review Once installed, you can modify the evaluation thresholds by directly editing the logic within skills/self-review/index.js to better suit your specific output standards.

Use Cases

This skill is perfect for developers building autonomous agents where output quality is non-negotiable. Use it to prevent "hallucinated" or "bloated" responses from reaching users. For example, in a technical support chatbot, the skill can ensure that every response provides a clear, actionable step. In creative writing agents, it can be configured to force conciseness, ensuring the generated text remains punchy and engaging. By integrating this into step 6 of your AGENTS.md workflow, you ensure a consistent quality bar across all interactions.

Example Prompts

1. "Review my previous response about the Python installation script and let me know if it meets the clarity and conciseness requirements."
2. "Run the self-review skill on the following documentation draft: [insert text here]."
3. "Filter my output through the self-review pipeline before submitting to the client; if the exit code is 1, summarize the errors for me."

Tips & Limitations

While the Self-Review skill is powerful for consistency, remember that it is rule-based, not context-aware. It excels at checking sentence length and structure, but it may miss nuances regarding factual accuracy or tone. For complex reasoning or fact-checking, consider pairing this with the self-review-llm package. Always remember to test your custom threshold settings on a sample dataset before deploying to production to ensure the skill doesn't become overly restrictive.