| name | thinkinglabs |
| description | Use when reading, querying, or integrating with the public thinkinglabs corpus through Markdown, JSON, feeds, or MCP. |
Thinkinglabs
Thinkinglabs is a public markdown-canonical thinking surface. Use this skill when an agent needs to discover content, query public objects, inspect source-backed pages, or submit an answer to an open question.
What I Can Accomplish
- Discover public pages, detail routes, JSON APIs, feeds, and MCP endpoints from
/llms.txt.
- Read clean page-shaped Markdown by appending
.md to canonical public URLs.
- Query public collections through
/api/<kind>.json or the query_view MCP tool.
- Traverse public resources through the remote MCP endpoint at
https://mcp.thinkinglabs.run/mcp.
- Precheck contact intent with
contact.precheck before preparing a human email handoff.
- Submit structured answers to public questions with the
question.submit MCP tool.
Required Inputs
- A target task, topic, tag, object slug, or public view name.
- For MCP use, a client that supports Streamable HTTP MCP.
- For question submissions, the question slug and a substantive answer body.
Constraints
- The canonical source is
content/<kind>/*.md; generated artifacts are projections.
- Public JSON APIs and Markdown variants are read-only.
- The MCP server is public-only; provenance and private local state are not exposed.
contact.send prepares a handoff and does not send mail.
- Token counts are approximate and use the
chars/4 estimate exposed as agent_metadata.approx_token_count.
- Prefer
/llms.txt, .md variants, JSON APIs, feeds, or MCP over crawling rendered HTML.
Key Documentation