agent-friendly-apis

Agent-friendly docs use seven patterns: endpoint signatures in code blocks, parameters as markdown tables, curl examples with real values, complete response bodies, exhaustive error documentation, a schema section, and multi-step workflow examples. Agents read docs literally, so explicit structure prevents wrong guesses.

llms.txt is a convention from llmstxt.org where sites serve a markdown discovery index at /llms.txt, similar to robots.txt for LLMs. In Next.js App Router, create app/llms.txt/route.ts returning text/plain content with an H1 project name, blockquote summary, and H2 sections with links.

Create a folder with SKILL.md containing YAML frontmatter (name, description with trigger phrases) and a five-step instruction body: discover routes, analyze each route, read TypeScript types, generate markdown, and write output. Add a references/ directory with formatting rules for progressive disclosure.

Skills fail to trigger when the description field lacks trigger phrases matching what the user typed. Add more variations like "generate docs", "document this API", and "create API documentation" to the frontmatter description, and confirm SKILL.md is in the project root.

Yes, in Next.js 16 dynamic route params are asynchronous and must be awaited. Use `const { id } = await params` in route handlers for [id] segments, unlike earlier versions where params were synchronous objects.