agent-docs

What This Skill Does

The agent-docs skill provides a standardized framework for authoring documentation specifically optimized for AI agent consumption. Instead of writing for human readers, this skill helps developers structure content for RAG retrieval, token efficiency, and the Hybrid Context Hierarchy. It enforces best practices such as inline context, primacy-bias awareness, and machine-readable formats like llms.txt, ensuring that agents can successfully retrieve and act upon project knowledge without hallucination or inefficient tool calling.

Installation

To install this skill, run the following command in your OpenClaw terminal: clawhub install openclaw/skills/tylervovan/agent-docs

Use Cases

• Project Onboarding: Creating an AGENTS.md file that acts as the source of truth for an agent, defining architecture constraints and security rules.
• Optimized Context Injection: Converting large, sprawling documentation into structured, token-efficient chunks that LLMs can process in their context window.
• Standardization: Ensuring all repositories within an organization follow the same hierarchical structure (Constitution -> Reference Library -> Research Assistant).
• Debugging: Reducing the 'Lost in the Middle' problem by ensuring critical project constraints are placed at the top of documentation files where agents are most likely to prioritize them.

Example Prompts

1. "Generate an AGENTS.md file for my current React project based on the Hybrid Context Hierarchy. Include security rules and a reference to the schema.md file."
2. "Refactor my existing README.md into an llms.txt-compliant format to improve agent retrieval accuracy."
3. "Analyze my current project documentation and identify sections that are too long for agent consumption, then suggest a compression strategy."

Tips & Limitations

• Primacy Bias: Always place the most critical information (security constraints, directory structure, core build commands) at the very top of your documents. Agents weigh these tokens significantly higher than those in the middle of a file.
• Signal-to-Noise: Strip away marketing jargon, fluff, and version history. The cleaner the content, the higher the agent success rate.
• Limitations: This skill focuses on the structure of documentation. It does not automatically update your documentation when code changes; you must integrate this into your CI/CD or documentation workflow.