← ClaudeAtlas

agent-discoverabilitylisted

Use when publishing or auditing an MCP or agent endpoint so third-party agents can find it, understand its capabilities, and complete its authorized authentication flow.
sandbaseai/workbuddy-skill · ★ 2 · AI & Automation · score 81
Install: claude install-skill sandbaseai/workbuddy-skill
# Agent Discoverability Use this Skill when a team owns an Agent/MCP endpoint and wants other agents and users to find and connect to it. This covers the publish/serve side, not driving a third party's MCP as a client and not ordinary search-engine SEO. ## Establish ownership and scope Confirm the exact endpoint, domain, transport, public tools, audience, authentication model, and operator authorization before changing anything. Inventory the canonical registry entry, community listings, deployment routes, well-known documents, DNS zone, certificate owner, and current versions. Never invent a domain-ownership proof, capability, OAuth URL, DNS record, or token. Use placeholders in examples and keep real credentials in the authorized secret store or environment. ## Keep the four discovery surfaces consistent An agent typically needs four surfaces. They may be delivered incrementally, but all values must describe the same endpoint: 1. **Registry listing:** canonical name, repository, version, transport, HTTPS remote, and capability/manifest URL. Mirror only after the canonical record is verified; record each listing's owner and update path. 2. **OAuth discovery:** an unauthenticated protected request returns `401` and a `WWW-Authenticate: Bearer` header with a `resource_metadata` URL; the protected-resource metadata names the resource and authorization server; the authorization-server metadata exposes the actual authorize/token endpoints. 3. **Capability man