Docs For Machines

Every page on this site answers to a browser. These endpoints answer to software. The catalog below is also published as data at /.well-known/api-catalog, and every page carries a Link header pointing at it, so nothing here has to be found by reading this page.

• /.well-known/api-catalog API catalog, as a linkset. Lists every API with its description and its docs.
• /openapi.json OpenAPI 3.1 description of the read only HTTP surface.
• /.well-known/mcp/server-card.json MCP server card. Draft format.
• /.well-known/agent-skills/index.json Agent Skills index, with a SHA256 digest for every skill.
• /auth.md How authentication works here. Short answer: it does not.
• /robots.txt robots.txt, including content signals.
• /feed.xml RSS 2.0 feed of the blog.

Send Accept: text/markdown and a content page returns markdown instead of a profile full of tables and blinking chrome. Responses carry Vary: Accept, so a shared cache keeps the two apart, and X-Markdown-Tokens carries an estimated token count for the body.

The same markdown is published at its own address, which never negotiates and therefore caches cleanly. Prefer these when you know what you want:

• /index.md is /. Profile, credentials, featured prompts, and latest blog entry for tostupidtooquit.com, rendered as markdown for agents.
• /prompts.md is /prompts. Every prompt in the library, with body, variables, and usage notes, as markdown.
• /blog.md is /blog. Every blog entry, newest first, as markdown.
• /about.md is /about. What tostupidtooquit.com is and who runs it, as markdown.

The token count is an estimate. Exact counts depend on the tokenizer, and this server has no way to know which one you are using.

/mcp speaks the Model Context Protocol over Streamable HTTP. POST a single JSON-RPC message. Protocol revisions supported: 2026-07-28, 2025-11-25, 2025-06-18. GET and DELETE return 405, because the current revision has no standalone stream and no sessions.

No credentials. No writes. Nothing here can send mail, post to the guestbook, or reach the intake form. The tools are:

• search_site: Full site keyword search across prompts, blog entries, pages, services, and playlists. Returns ranked results with a URL for each. Use this first when you do not already know an identifier.
• list_prompts: List prompts in the free library, optionally filtered by category. Returns identifiers and summaries, not bodies. Call get_prompt for the pasteable text.
• get_prompt: Fetch one prompt by identifier, including the full body exactly as the site's copy button would put it on the clipboard, plus its placeholders, usage notes, and tips.
• list_blog_posts: List every blog entry, newest first, with identifier, date, category, and teaser.
• get_blog_post: Fetch the full text of one blog entry by identifier.

The same tools are offered to in browser agents through WebMCP, which is a draft browser API. When a browser exposes it, this site registers the list above; every call goes back to /mcp, so both paths return identical answers.

The prompts filed under Agent Skill are also published as skills, one SKILL.md each, indexed at /.well-known/agent-skills/index.json. Every entry carries a SHA256 digest of the file it points at, computed at build time from the bytes actually served, so a client can verify what it downloaded and skip a fetch when nothing changed.

• requirement-analysis-and-planning-agent, from p45
• social-media-post-analyzer, from p46
• ticket-to-pr, from p47
• app-feature-focused-readiness-audit, from p48
• data-lineage-agent-skill, from p49
• sniper-precision-debugging-skill, from p50
• add-ai-protection, from p51
• borrow-skill, from p52
• task-creator, from p53
• trello-integration-skill, from p54
• codebase-onboarding-skill, from p71
• changelog-generator-skill, from p72
• research-brief-skill, from p73
• pull-request-description-skill, from p74

/robots.txt carries a Content-Signal line. Search and AI input are granted: the prompts are free, and answering somebody's question with them is the entire point. Training is left unstated, which under the published policy neither grants nor refuses it.

Content signals say what may be done with content after it is fetched. They are not access control. Allow: / still governs who may crawl.

Link headers follow RFC 8288 and use only relation types in the IANA registry. The catalog follows RFC 9727 and serializes as RFC 9264 linkset. Content negotiation follows RFC 9110. Well known paths follow RFC 8615. The MCP server card and WebMCP are drafts and are labelled as such wherever they appear.

A full mapping of every endpoint to the document that defines it lives in the repository under docs/agent-discovery/.