Does the AI Skill Store API require authentication?

Read-only endpoints (skill discovery, search, schema, download) require no authentication. Upload and management endpoints require Authorization: Bearer in the request header.

How do agents use the AI Skill Store API?

Agents start at GET /llms.txt or GET /v1/agent/info for a machine-readable overview. Then search by capability with GET /v1/agent/search?capability= , fetch the call schema with GET /v1/agent/skills/{id}/schema, and download a platform-specific package with GET /v1/agent/skills/{id}/download?platform=ClaudeCode.

What is the difference between /v1/skills/* and /v1/agent/* endpoints?

/v1/skills/* endpoints are for human developers — they include upload, versioning, reviews, and management. /v1/agent/* endpoints are optimized for programmatic agent access — they return structured JSON with schemas, capabilities, and pre-built download packages.

📡 API Reference

REST API · OpenAPI 3.0 · Agent-First

llms.txt ↗ apispec.json ↗

Auth

Read endpoints: no auth
Upload: Authorization: Bearer <api_key>

Agent API

/v1/agent/search · /v1/agent/skills/{id}/schema · /v1/agent/skills/{id}/download

Platforms

Auto-converts USK v3 skills for OpenClaw · ClaudeCode · CustomAgent

Discovery

Agents start at /llms.txt or /v1/agent/info