Rtfm Testing

What This Skill Does

RTFM Testing is an innovative documentation quality methodology designed to eliminate the 'curse of knowledge' inherent in technical writing. Created by zscole, this skill automates the process of validating documentation by deploying 'blind' AI agents—entities with zero prior context—to execute tasks based solely on your provided guides. By simulating a fresh user experience, the skill systematically exposes unconscious assumptions, skipped steps, and missing prerequisites that human authors often overlook. The process transforms documentation from a static asset into a testable component of your development lifecycle, ensuring that if an agent can finish the task using the docs, a real user likely can as well.

Installation

To integrate this documentation validation tool into your environment, run the following command within your terminal or OpenClaw interface: clawhub install openclaw/skills/skills/zscole/rtfm-testing. Ensure you have the GAPS.md and TESTER.md files available in your repository root, as these define the output format and behavioral constraints required for the tester agent to function correctly.

Use Cases

RTFM Testing is best utilized during critical points of the documentation lifecycle. Use it before public releases to perform a 'sanity check' on setup guides, or integrate it into your Continuous Integration (CI) pipeline for documentation-heavy projects to ensure that major refactors do not break existing tutorials. It is also an effective troubleshooting tool when user feedback indicates confusion that your team cannot reproduce. By treating every confusion point as a technical bug, you create a rigorous feedback loop that elevates the quality of your developer experience.

Example Prompts

1. "Run an RTFM test on the new OAuth configuration guide. Use the TESTER.md prompt and ensure the agent starts with zero prior knowledge of our API."
2. "I need to validate our installation readme. Please spawn a tester session using the docs in docs/setup.md and report all gap categories in GAPS.md format."
3. "The users are reporting errors in the deployment tutorial. Execute an RTFM test using the provided docs and tell me specifically where the agent gets stuck."

Tips & Limitations

To get the most out of RTFM Testing, adhere strictly to the principle of 'no hints.' The AI tester should not be allowed to use common sense or external search capabilities; it must rely solely on the text provided. If the agent fails, do not assist it—analyze the point of failure as a documentation deficiency. The main limitation is that the effectiveness depends heavily on the quality of your TESTER.md system prompt; ensure it is configured to be strictly literal to avoid false positives. Remember: if the documentation is too sparse for an AI to follow, it is definitely too sparse for a human user.