Skip to content
Reference

Agent skills

Everything this site exposes so AI agents can discover, read, and interact with Atelier UI automatically.

Discovery endpoints

All well-known well-known file A URL at a predictable path (usually under /.well-known/…) that machines check for metadata, following RFC 8615. Examples: /.well-known/security.txt, /.well-known/mcp/server-card.json. and machine-readable files published at this domain:

PathContent-TypePurpose
/.well-known/api-catalog application/linkset+json RFC 9727 API catalog — lists MCP endpoints
/.well-known/mcp/server-card.json application/json SEP-1649 MCP Server Card
/.well-known/agent-skills/index.json application/json Agent Skills RFC v0.2.0 index
/robots.txt text/plain Content-Signal directives (ai-train, search, ai-input)
/llms.txt text/plain Short component API index (llms.txt standard)
/llms-full.txt text/plain Full component API reference with all props and examples

HTTP agent features

A

Link Headers (RFC 8288)

The homepage response includes a Link header advertising the component API reference, full docs, and the API catalog — so agents discover resources from the first HTTP response without crawling.

Link: </llms.txt>; rel="describedby",
</llms-full.txt>; rel="service-doc",
</.well-known/api-catalog>; rel="api-catalog"
B

Markdown for Agents

Any page can be requested as plain Markdown by sending Accept: text/markdown. The response will have Content-Type: text/markdown and an x-markdown-tokens count — no HTML parsing needed.

Terminal window
curl -H "Accept: text/markdown" https://atelier.pieper.io/workshop
C

Content signals

robots.txt declares AI content usage preferences per the Content Signals draft standard: agents may use this content as inference input but not for model training.

# Content Signals — https://contentsignals.org/
Content-Signal: ai-train=no, search=yes, ai-input=yes

Workspace skills

Multi-file Claude Code skills that ship architectural playbooks — references and assets loaded on demand by the active mode. Distinct from the MCP Skills below: these run in Claude Code itself, not as a remote MCP server.

α

figma-workspace-architect

Designs, builds, audits, and migrates Figma workspaces via figma-console-mcp. Four modes (Build, Audit, Decide, Migrate) with three sub-modes (Scaffold, Re-verify, Code-verify). 13 references loaded on demand.

https://atelier.pieper.io/.well-known/agent-skills/figma-workspace-architect/SKILL.md

MCP skills

Three Storybook MCP servers — one per framework — expose component documentation via the Model Context Protocol. Each skill is described in a SKILL.md file listed in the agent-skills index.

01

Angular — storybook-angular

MCP endpoint: https://atelier.pieper.io/storybook-angular/mcp

Available tools: list-all-documentation, get-documentation, get-documentation-for-story

https://atelier.pieper.io/.well-known/agent-skills/storybook-angular/SKILL.md
02

React — storybook-react

MCP endpoint: https://atelier.pieper.io/storybook-react/mcp

Available tools: list-all-documentation, get-documentation, get-documentation-for-story

https://atelier.pieper.io/.well-known/agent-skills/storybook-react/SKILL.md
03

Vue — storybook-vue

MCP endpoint: https://atelier.pieper.io/storybook-vue/mcp

Available tools: list-all-documentation, get-documentation, get-documentation-for-story

https://atelier.pieper.io/.well-known/agent-skills/storybook-vue/SKILL.md