By name: agents-md-gen description: Generate valid AGENTS.md files for repositories according to AAIF standards. Use when Claude needs to create or update AGENTS.md files that document AI agents in a repository following standard format and best practices.
AGENTS.md Generator
This skill generates AGENTS.md files that document AI agents in a repository according to AAIF (AI Agent Interoperability Framework) standards. It analyzes a repository's structure, identifies agent configurations and conventions, and outputs a properly formatted AGENTS.md that makes the repo AI-tool-friendly.
Use this skill any time you set up a new agent-enabled project or onboard an existing repo to the AAIF standard.
Quick Start
# From the repo root, invoke the skill
/agents-md-gen
# The skill will scan your repo and write AGENTS.md
The output file is placed at the repository root as AGENTS.md.
Key Features
- Scans repository structure for agent-related files (
.claude/,skills/,agents/, config YAMLs) - Identifies and documents available tools, permissions, and memory configuration
- Outputs AAIF-compliant AGENTS.md with correct section headings
- Adds repository conventions (coding style, commit format, test commands)
- Idempotent — re-running updates rather than overwrites existing content
Process
- Analyze the repository structure
- Identify agent-related files and configurations
- Document repository conventions and guidelines
- Generate a properly formatted AGENTS.md file
The generation logic is implemented in the supporting script to maintain consistency and follow AAIF standards.
When NOT to Use This Skill
- Non-agent repositories — if there are no AI agents or Claude Code integrations, an AGENTS.md adds noise without value
- Already documented projects — if a complete, accurate AGENTS.md exists, editing it manually is faster than regenerating
- Proprietary internal tooling — if your agent configuration contains secrets, verify the generated file before committing
Common Mistakes
- Running the skill in a subdirectory instead of the repo root — the scanner misses top-level config files
- Not reviewing the generated file before committing — always check agent permissions and memory settings are correct
- Leaving placeholder tool descriptions — fill in real examples so other AI tools can understand the agent's capabilities
Related Skills
mcp-builder— Build the MCP servers that AGENTS.md documentsimplementation-specialist— Implement the agent workflows referenced in AGENTS.mdaudit-logging-system— Add audit logging to the agents documented here