Skills — reusable playbooks for AI agents

A skill is a step-by-step playbook an AI agent can follow end-to-end. Where the MCP tools are the individual actions an agent can take, a skill is the plan that strings them together for a whole job — for example, importing an existing website into StaticOwl as a self-hosted, structured, editable site.

Skills are maintained by StaticOwl and ship as part of the product. They're the same for every site, and they're read-only: the dashboard, the API, and the MCP tools all show the same catalog — none of them creates or edits a skill. There's no per-site authoring step.


Where skills show up

Surface What you get
Dashboard → Skills Browse the catalog, open a skill to read it, and copy the full playbook.
MCP (skills_list, skills_get) An agent discovers and loads a skill while it works.
Public API (/api/skills) The same catalog over HTTP — no authentication needed.

Use a skill with an AI assistant

The point of a skill is to hand an agent a working plan. With the StaticOwl MCP connected, an agent:

  1. Calls skills_list to discover what playbooks are available.
  2. Calls skills_get with a skill's name to load its full instructions.
  3. Follows the playbook, using the other MCP tools to do the work.

If you're driving the agent yourself, you can also open a skill in the dashboard, Copy playbook, and paste it into a chat with the StaticOwl MCP connected. Or, for Claude Code, drop the skills/<name> folder into .claude/skills/ and invoke it as /<name>.

See Connect your AI assistant for how to connect the MCP.


What's in the catalog

Today the catalog includes:

The list grows over time; call skills_list or check the dashboard for the current set.


API and MCP reference

Both surfaces are read-only.

Route MCP tool Returns
GET /api/skills skills_list A summary of each skill: name, title, description, updatedAt.
GET /api/skills/{name} skills_get One skill, including its full Markdown playbook.

The /api/skills routes are public — no API key required — because a skill is documentation meant to be discovered and shared.


See also

Related docs

Resolved at build time via {% similar %} — cosine similarity over embeddings, not tag overlap. Zero arguments.

Connect your AI assistant
69% match
Zero to deployed — the 5-minute quickstart
62% match
AI features
59% match