A collection of public API documentation converted to Agent Skills format - structured markdown optimized for AI agents.

AI agents often need to interact with external APIs. Raw OpenAPI specifications have limitations:

• Too large - Complex specs can be megabytes, exceeding LLM context limits
• Wasteful - Loading the entire document for every request wastes context
• Monolithic - All-or-nothing access prevents selective loading

Agent Skills solves this by splitting documentation into on-demand files. Agents load only what they need - start with an overview, drill into specific operations.

No special integrations required. File reading works in every agent framework.

Point your AI agent to read apis/<api-name>/<skill-name>/SKILL.md as the entry point.

Example for Stripe:

apis/stripe/stripe-api/SKILL.md

Each API is converted using openapi-to-skills:

npx openapi-to-skills ./openapi.json -o ./output

The tool transforms monolithic OpenAPI specs into a navigable file structure:

<skill-name>/
├── SKILL.md                    # Entry point with overview
└── references/
    ├── resources/              # One file per resource/tag
    ├── operations/             # One file per operation
    ├── schemas/                # Grouped schema files
    └── authentication.md

Want to add an API? Requirements:

1. Public API with an OpenAPI 3.x specification
2. Permissive license allowing redistribution of documentation

To add:

# Create directory
mkdir -p apis/<api-name>

# Download OpenAPI spec
curl -o apis/<api-name>/openapi.json <spec-url>

# Generate skills
bunx openapi-to-skills apis/<api-name>/openapi.json -o apis/<api-name>

# Add README and update this file

Each API's documentation retains its original license. See individual API directories for details.

This repository structure and tooling is Apache-2.0.