A platform team wants its MCP server to share reusable agent workflows, such as a code-review procedure with a checklist file, that hosts can discover and load on demand. What does the Skills over MCP extension provide?
Select an answer to reveal the explanation.
Short Explanation
Think of it as a shared recipe binder: your host browses the index, pulls the recipe card when it's needed, and reads the attached notes. The trap is picturing skills as tools that run on the server.
Full Explanation
The Skills over MCP extension, identified as io.modelcontextprotocol/skills, lets servers expose workflow instructions and supporting files to clients. A skill is a directory containing a SKILL.md file with name and description frontmatter plus optional supporting files. Servers that declare the extension implement skills/list and skills/get to provide skill entries with manifests, and the files themselves are read through resources/read, so the extension reuses the existing resources primitive rather than adding a new execution path. Hosts decide whether the model or the user selects a skill, verify file sizes and digests against the manifest, and apply any required user approval before loading instructions into context. It is not a new invocable primitive that runs workflows server-side; reading SKILL.md does not execute anything, and actions still go through tools. Nothing is pushed into conversations at connection time, and the extension explicitly forbids retrieving files ahead of need. Elicitation collects user input during a request and is unrelated to distributing reusable instructions. Exam caveat: skill content is untrusted input, so hosts must tag it with its originating server and require explicit approval for permission grants. Operational check: confirm server/discover declares both the resources capability and the skills extension, then load one skill and verify its digests.