Imported from jessenaiman/old-macdonald-had-a-school (AGENTS.md). Install upstream with npx skills add jessenaiman/old-macdonald-had-a-school. Copyright stays with the author.

1. Read before answer. Library/API question → Context7 MCP (required external; not in project .mcp.json): resolve-library-id → query-docs → cite the fetched snippet. No memory-answers about this project's packages.
2. Next.js authority = installed version. Use the project-registered next-devtools MCP: nextjs_docs first (it answers use_bundled_docs + docsPath), then read/grep node_modules/next/dist/docs/. Live-server checks (nextjs_index / nextjs_call) only after npm run dev is actually up.
3. Skills are files, not commands. Before acting under a skill, read its .agents/skills//SKILL.md and follow exactly what it specifies — e.g. impeccable: run its scripts/context.mjs once per session (--target <path>, cwd at project, never rerun), load the reference/*.md playbook for the sub-command asked (critique, audit, polish, harden, clarify, live); for drift use the documented $impeccable doctor after reading reference/doctor.md. Relevant here: building-components, impeccable, next-dev-loop, verify-and-stop, ponytail-audit, agent-browser.
   1. Skill = markdown instruction file, not a command. Discovery lists name+description in prompt at startup.
   2. Use it: read skill:// → obey what it says → run its named commands via real tools (bash node .../context.mjs, read skill://impeccable/reference/critique.md, browser, etc.). Sub-paths resolve inside the skill dir: skill:///.
   3. Matching: request matches description = trigger → read skill before acting. That's the MUST.
   4. User-side: /skill: (e.g. /caveman ultra) injects the body; skills persist until "off".
4. UI claims need pixels. Browser tool (skill://agent-browser for the workflow): open the route, screenshot desktop 1280 + mobile 375, light + dark. "Verified" names the screenshot path.
5. User deletions stay deleted. NEVER git checkout/restore on uncommitted files unless the user names those paths. Commit deletions when asked; never "repair" them.
6. User-pasted URL → read it first, quote it, then verdict. Unread = no opinion.
7. Plan is mandatory state: read the active plan file and check the todo list at every resume; execute to completion; never yield mid-plan.
8. Gates before commits: npm run typecheck && npm run lint && npm run build → exit 0. Failures verbatim.
9. Coding/check-in: before any coding task, read docs/superpowers/check-in-workflow.md; use one short-lived branch/PR, push only after the local commit gate and task checks pass, and never merge without owner approval. At integration, resolve/read skill://finishing-a-development-branch and follow it exactly.
10. Chat: caveman terse, no essays, no inventory recitals. Files/commits: normal prose.

This is NOT the Next.js you know

This version has breaking changes — APIs, conventions, and file structure may all differ from your training data. Read the relevant guide in node_modules/next/dist/docs/ (resolved from this file's directory; in monorepos the next package may not be visible from the repo root) before writing any code. Heed deprecation notices.

This block is written and re-added by next dev — verify at node_modules/next/dist/server/lib/generate-agent-files.js. Removing it from a diff only re-creates the uncommitted change; committing it with your work keeps the tree clean.