# Em 0022

- Object ID: `em:task:sha256:d529afb55de3ec5e86773d723762eb4c60eeb16576e287003840eba8bf7ed9b4`
- Kind: `task`
- Repository path: [`tasks/contracts/EM-0022.json`](https://github.com/yoheinakajima/epistemedia/blob/f92846570180dfa4511263f8ba98ecd18f7772c9/tasks/contracts/EM-0022.json)
- Content digest: `45bd5d82e823468b22e9e764a7adb10d72f234d702628c88c3232fed1debf8f8`

**Also filed under:** [Agent Operations](https://epistemedia.org/topics/agent-operations/), [Research Program](https://epistemedia.org/topics/research-program/)

## Source content

{
  "$schema": "https://epistemedia.com/schemas/task-contract-v1.json",
  "id": "EM-0022",
  "title": "Harden the doors, review trust, and agent reachability for Case 001",
  "status": "ready",
  "change_class": "ordinary-implementation",
  "objective": "Make every primary human and agent entry route lead coherently into the accepted Case 001 evidence experience, replace the unlinked independent-review badge with a bounded public review receipt, expose one first-screen path into the unresolved lineage, and encode deterministic reachability and release-identity parity checks without changing the accepted research dossier.",
  "depends_on": [
    "EM-0021"
  ],
  "authority": {
    "allowed_paths": [
      "README.md",
      "catalog/**",
      "src/**",
      "tests/**",
      "docs/**",
      "ops/**",
      "runs/**"
    ],
    "forbidden_paths": [
      "constitution/**",
      "policies/**",
      "schemas/**",
      "governance/events/**",
      ".github/**",
      "research/**",
      "tasks/contracts/**"
    ]
  },
  "required_evaluation": [
    "How We Know, Substrate, Case 001, skeptical, Markdown, JSON, llms.txt, robots.txt, and sitemap route tests",
    "sanitized public review-receipt scope, reviewer, exact-head, and limitation tests",
    "cold-start agent discovery and Case 001 summary test using only committed public outputs",
    "whole-release catalog, frontier, compiler, accepted-commit, and manifest identity-parity test",
    "no-JavaScript fallback and keyboard-accessibility review",
    "desktop and mobile visual inspection",
    "public disclosure noninterference audit",
    "external bot-user-agent, route, content-type, canonical, and release-identity read-back after accepted deployment"
  ],
  "acceptance": [
    "the How We Know route is a compact section index rather than a duplicate homepage",
    "the existing repository-internals Explore destination is accurately labeled Substrate in primary navigation",
    "the Case 001 review label links to a public receipt naming the reviewer, exact reviewed commit, checked scope, independence conditions, and limitations",
    "a first-screen evidence-count control opens the unresolved 2007 lineage without requiring a long scroll",
    "a cold-start agent can discover stable Markdown and JSON twins from llms.txt and recover the case question, policy-relative verdict, evidence counts, and unresolved lineage",
    "all generated public routes expose one coherent release identity and deterministic verification fails on mixed-version output",
    "the accepted Case 001 dossier, independent receipt, and research packet remain byte-identical",
    "make check passes without changing accepted source state",
    "the accepted Pages deployment passes human, bot, canonical URL, keyboard, mobile, content-type, and release-identity read-back"
  ],
  "limitations": [
    "This task does not add, revise, or re-review Case 001 research claims, evidence spans, counts, lineage judgments, or policy conclusions.",
    "A deterministic cold-start agent fixture establishes machine-interface reachability and parity, not general comprehension by all agents.",
    "The Substrate label preserves the existing Explore route and self-describing repository corpus; it does not remove internals from public access.",
    "Case 002, additional lenses, hosted API/MCP activation, workflow mutation, DNS changes, publication, and release are out of scope."
  ]
}