Skills.sh: A community directory/registry for WebMCP tools (discovery use case)

[JustineDevs]

Summary

I've built Webmcp-skills, skills.sh a community project exploring a directory/registry model for WebMCP "skills" (tools). I'd like to share it here as it relates directly to the open discovery topic discussed in the explainer and proposal.

Relation to WebMCP

The current WebMCP design registers tools dynamically via navigator.modelContext.provideContext() / registerTool(), which requires an agent to navigate to a page and execute JavaScript before it can discover what tools are available. The proposal explicitly calls this out as a limitation:

"Tool discoverability: There is no built-in mechanism for client applications to discover which sites provide callable tools without visiting or querying them directly. Search engines, or directories of some kind may play a role in helping client applications determine whether a site has relevant tools for the task it is trying to perform."

Skills.sh is an attempt at exactly this — a directory that indexes and surfaces WebMCP-compatible tools/skills so agents (or developers) can discover relevant sites/capabilities without needing to visit each one first.

What I'd like feedback on

• Whether a directory/registry approach like this aligns with how the WG is thinking about discovery mechanisms (mentioned as a "Future Exploration" / open topic in the README and proposal).
• Whether there's interest in a lightweight manifest field (similar to the Web App Manifest "tools" idea discussed in docs/proposal.md and docs/service-workers.md) that a directory like Skills.sh could consume to index sites more reliably.
• Any existing discussions/issues I should align with (I noticed discovery is referenced near issue Declarative API Equivalent #22 regarding declarative tool definitions).

Links

• Project: https://github.com/JustineDevs/webmcp-skills
• Related docs: docs/proposal.md (Alternatives Considered / Tool discoverability), README.md (Open Topics / Future Explorations)

Happy to get feedback or contribute this work back into discovery-related discussions for the spec.