{
 "schema_version": 1,
 "path": "/caches/about/",
 "canonical": "https://agentsite.network/caches/about/",
 "kind": "about",
 "title": "About this directory",
 "description": "Sources, eligibility, classification limits and machine-readable formats of the Agentsite Network directory.",
 "breadcrumbs": [
  {
   "title": "Directory",
   "path": "/caches/"
  }
 ],
 "notes": [],
 "sections": [
  {
   "heading": "Inventory",
   "intro": "One entry per domain for every completed, published and not unpublished Agentsite. Publication evidence is a release, a completed extraction, or a legacy record that reports pages. Unknown first-publication dates are left blank."
  },
  {
   "heading": "Industries",
   "intro": "NAICS 2022 sectors and subsectors, plus two project classes (X-SW, X-PLAT) that roll up to sector 51 labels are AI-generated, unverified. AI-generated labels from each Agentsite's own page titles and descriptions, using a two-step NAICS rubric (sector, then subsector). Not human-verified. Confidence >= 0.55 is labelled classified; below 0.55 the top-1 class is labelled unconfirmed. Only classified labels at confidence >= 0.70 are published; other sites appear under All sites only. Parked domains and near-empty mirrors are not labelled. Source: EVE-611 rubric, full-inventory run, scripts/directory/labels/ (fetch-evidence, classify, export) at ec1662c3, generated 2026-10-28. Industry coverage does not decide listing: every eligible site is under All sites."
  },
  {
   "heading": "Curated topics",
   "intro": "Curated topics and company profiles carried over from the EVE-250 prototype. Descriptions are one-line summaries reviewed against each Agentsite's llms.txt (source path, relative to the Agentsite). Deep links are listed only when a content-aware check passed (HTTP 200, HTML, at least 80 words of text)."
  },
  {
   "heading": "Lists",
   "intro": "Every list holds at most 100 sites and is split by domain prefix. List URLs stay valid when a list is split. Navigation needs no JavaScript, search, cookies or login."
  },
  {
   "heading": "Formats",
   "intro": "Each HTML page has Markdown (index.md) and JSON (index.json) twins built from the same records. JSON pages use schema_version 1: path, kind, title, sections[].links[] {title, path or url, count} and sections[].sites[] {domain, url, pages, updated}.",
   "links": [
    {
     "title": "Network summary (JSON)",
     "path": "/caches/network.json"
    },
    {
     "title": "Guide for language models (llms.txt)",
     "path": "/caches/llms.txt"
    },
    {
     "title": "Directory sitemap index",
     "path": "/caches/sitemaps/index.xml"
    }
   ]
  }
 ]
}
