qa-gate

What This Skill Does

The qa-gate skill serves as the ultimate quality assurance layer for the OpenClaw agent ecosystem. Designed as a non-negotiable release gate, it ensures that any artifact—ranging from complex PRDs and code implementations to blog posts and technical documentation—is rigorously vetted before reaching human eyes. Rather than acting as a code review tool, qa-gate operates as a final, high-fidelity checkpoint that evaluates content across six critical dimensions: Factual Accuracy, Tone Consistency, Completeness, Structural Integrity, Operational Soundness, and Sensitive Data Handling. By enforcing these standards, it guarantees that all outputs presented to the principal are professional, secure, and ready for publication.

Installation

To integrate this skill into your workflow, execute the following command in your terminal: clawhub install openclaw/skills/skills/bloodandeath/keats-qa-gate

Use Cases

• Pre-Publication Review: Use this before pushing content to ClawHub or Gumroad to ensure high quality.
• PRD Validation: Automatically run this after any PRD generation loop to catch missing sections or placeholders.
• Technical Documentation: Verify that code blocks, command sequences, and API references are valid and error-free.
• Sensitive Content Scrubbing: Use the sensitive data check feature to ensure no private keys, tokens, or personal identifiers are accidentally included in public-facing artifacts.
• High-Stakes Projects: Utilize the optional --dual flag to perform a redundant, cross-model verification for critical business documents.

Example Prompts

1. "QA gate this draft PRD for the new marketing engine and let me know if it's ready to share with the team."
2. "Run a final quality validation on this blog post. Make sure the tone is consistent and check for any placeholder text."
3. "Please perform a quality gate on these technical instructions. I need to be 100% sure the API commands are correct and no sensitive data is present before I release this to our users."

Tips & Limitations

• Do not skim: The skill requires the full document to function effectively; ensure the agent has access to the complete file content.
• Use --dual wisely: While highly accurate, the dual-model mode increases token latency and cost. Reserve it for high-stakes content.
• Not a Code Reviewer: While it validates the operational soundness of commands, it does not replace a deep-dive security or performance audit for complex codebases.
• Iterate: If the gate returns a FAIL, treat the output as a checklist. Fix all CRITICAL and MAJOR issues before attempting to present the work to the principal.