Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Guides writing documentation structured for AI agent consumption, covering token efficiency, RAG chunking, and context hierarchy. Applies to AGENTS.md constitutions, llms.txt indexes, SKILL.md files, and API docs read by LLMs. Based on Vercel benchmarks showing inline docs achieve 100% agent task pass rates versus 53% with retrieval-only approaches.
Inline AGENTS.md achieves a 100% agent task pass rate compared to 53% for tool-based retrieval alone, because agents skip fetching docs they assume they already know from training.
When to use it
- Writing AGENTS.md constitutions for a new codebase
- Generating a llms.txt index for an existing project
- Structuring API docs so an agent retrieves the right chunk
- Auditing a README for token efficiency before adding it to agent context
- Creating SKILL.md files for OpenClaw agent workflows
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Create an AGENTS.md for my Next.js project
- 1Identifies stack, architecture constraints, and critical governance rules
- 2Places security rules and build commands at the top to exploit primacy bias
- 3Writes a compressed docs index with file paths and function signatures, not full code
- 4Structures each H2 section to be self-contained for RAG chunk boundaries
- 5Validates total inline context stays under 4,000 tokens
AGENTS.md with governance rules, a docs index, and architecture constraints ready for agent context windows