# Em 0039

- Object ID: `em:task:sha256:b3f0b67206db72398fbf5797f993aaffe83c63989ea8f3e5dda444414da44433`
- Kind: `task`
- Repository path: [`tasks/contracts/EM-0039.json`](https://github.com/yoheinakajima/epistemedia/blob/f92846570180dfa4511263f8ba98ecd18f7772c9/tasks/contracts/EM-0039.json)
- Content digest: `256abb58d70965ab58ea97a7e450a48ac57d432b195e22e0f86bd8ca5145565b`

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

## Source content

{
  "id": "EM-0039",
  "title": "Publish the mission and four-case narrative",
  "status": "ready",
  "change_class": "ordinary-implementation",
  "objective": "Turn the four accepted How We Know cases into a coherent public introduction to Epistemedia: publish the owner-approved mission as a versioned record and deterministic About surface; explain the four distinct failure modes without changing any accepted dossier meaning; separate primary reader navigation from utility navigation; preserve honest current, live, and future capability boundaries; and publish a real-reader comprehension instrument whose completion requires actual human responses rather than synthetic inference.",
  "depends_on": [
    "EM-0037"
  ],
  "authority": {
    "allowed_paths": [
      "README.md",
      "catalog/**",
      "src/**",
      "tests/**",
      "docs/**",
      "ops/**",
      "runs/**"
    ],
    "forbidden_paths": [
      "constitution/**",
      "policies/**",
      "schemas/**",
      "governance/events/**",
      ".github/**",
      "tasks/contracts/**",
      "research/**",
      "catalog/dossiers/**"
    ]
  },
  "required_evaluation": [
    "byte and semantic preservation of every accepted dossier, independent review receipt, feature manifest, relation, count, evaluation, and evidence conclusion for Cases 001-004",
    "line-by-line editorial comparison with the owner-approved mission language while preserving the distinction among truth, evidence, bounded policy-relative evaluation, provenance, and current implementation state",
    "deterministic HTML, Markdown, and JSON parity for the About surface, including canonical URLs, stable content identities, discovery, sitemap, llms.txt, local REST, MCP, and CLI exposure where those read-only interfaces already apply",
    "homepage first-viewport review confirming that the featured evidence case remains substantive and the site does not become a broad marketing manifesto",
    "four-case narrative review confirming that overgeneralization, false independence, missing comparison class, and scope inflation are editorial failure-mode labels rather than stored evidence statuses or universal verdicts",
    "primary versus utility navigation review at desktop and mobile widths with no JavaScript, native keyboard access, visible focus, no overflow, canonical internal links, and stable agent discovery",
    "fail-closed regression tests for current live capabilities and explicitly future capabilities, including hosted API, hosted MCP, authenticated proposal submission, admission, and automatic merging",
    "real-reader comprehension instrument review confirming that synthetic tests cover mechanics only and that no human-comprehension result is claimed without retained answers from an actual reader",
    "public disclosure noninterference audit, whole-release identity parity, exact path-scope comparison, full deterministic make check, clean tracked state, and fresh-clone exact-head independent review",
    "protected merge, resulting-main validation, separately authorized Pages deployment, and live custom-domain route and release-identity read-back before any claim that the new public surface is live"
  ],
  "acceptance": [
    "a versioned mission record preserves the owner-approved voice while accurately stating that public readings are compiled from accepted records under named policies rather than stamping global truth",
    "the mission is available at stable /about/ HTML, Markdown, and JSON routes with a visible version, governance boundary, current-state section, and machine-readable identity",
    "the homepage contains a compact purpose statement, the featured case, a four-case failure-mode narrative, a three-step explanation of how Epistemedia reads a claim, and participation paths without replacing the forensic editorial hierarchy",
    "the How We Know index presents Cases 001-004 as four unit tests for how knowledge can fail, with each case's familiar claim, failure mode, bounded finding, defining count, and unresolved boundary",
    "primary navigation exposes How We Know, About, and For agents while Substrate, Docs, Status, and GitHub remain available as visibly secondary utilities",
    "current copy states that the four-case library and cold-start research kit are live only after verified deployment, while the hosted API, hosted MCP write surface, and authenticated submission queue remain unavailable until separately governed and deployed",
    "a public reader-check instrument asks a human to explain the evidence unit, run-independence rule, lens difference, a current limitation, and a source-inspection path; its result remains pending until an actual human responds",
    "all accepted Case 001-004 evidence artifacts and meanings remain unchanged and make check passes without changing accepted source state"
  ],
  "limitations": [
    "This task changes mission, information architecture, navigation, and explanatory presentation; it does not change accepted evidence, source spans, counts, policy semantics, or case evaluations.",
    "The four failure-mode labels are an editorial map across the first library, not a complete epistemology taxonomy or a new evidence classification.",
    "This task may describe and link the non-admitting research kit but may not implement EM-0038, accept hosted submissions, admit evidence, create or merge contribution pull requests, or expose a write-capable API or MCP service.",
    "No synthetic agent, browser test, or model review may be reported as a real-reader comprehension result.",
    "No research provider call, dossier research, policy change, constitution change, credential use, account provisioning, spend, DNS change, publication, or deployment is authorized merely by registering this task; protected merge and Pages deployment require their normal later gates."
  ]
}