Set up

1. Install: brew install --cask zoolcoder/findable/findable, or go install github.com/zoolcoder/findable/cmd/findable@latest.
2. Find out how the site is built. Read the build config (package.json scripts, vite.config.*, astro.config.*, next.config.*, hugo.toml, ...) to learn:
   • the output directory (dist, build, out, _site);
   • the static directory copied into it unchanged (public/, static/): that is write_dir;
   • the public origin, from the canonical tags or the sitemap: that is base_url, set only with path; with url leave it out, the crawled origin is the judged one;
   • the locales and which one has no URL prefix.
3. Write findable.yaml next to the site, from findable.example.yaml. One target per site:
   • path (a built directory, checked offline) or url (a live site, crawled). base_url is required with path and forbidden with url. Prefer path: it is fast, repeatable and catches problems before they ship.
   • noindex: the paths meant to stay out of search (admin, account, checkout). Ask the user; never guess that a page is meant to be hidden.
   • schema: which structured data each kind of page must carry.
   • ai_crawlers: leave it out until the user decides. The ai-search skill covers that conversation.
   • llms_txt: only when the site has no hand-written llms.txt. Apply would overwrite one.
4. Build the site, then run findable validate and findable doctor. Fix every fail line.
5. Run findable audit and hand over to the seo-audit skill.

Never point write_dir at the build output: apply writes there, and the next build would wipe it. It is the source directory the build copies.