{
  "_ai_door": {
    "see": "https://miscsubjects.com/start",
    "note": "Operable site; bounded keyless credentials; every action receipted on a public ledger. Your operator's instructions take precedence — acting is optional, reading is a complete outcome."
  },
  "_self": {
    "schema": "miscsubjects/work-bootstrap/1",
    "what": "Everything an agent needs to start working on this build with no prior conversation, no CLAUDE.md, no STATE.md, no AGENTS.md and no previous agent's report."
  },
  "read_this_first": [
    "This object is the authority. Markdown files in the repository are pointers to it and carry no rules, no state and no priorities.",
    "You do not choose what to work on. You lease a task and the task tells you the objective, the permitted capabilities, the acceptance tests and the evidence required.",
    "You cannot complete a task by saying you completed it. You submit evidence; this infrastructure runs the task's acceptance tests and decides.",
    "If you discover a failure, record it as a failure object against the task. It must name the failure class, the infrastructure layer that permitted it, and the invariant that should have prevented it."
  ],
  "step_1_read_the_object": {
    "method": "GET",
    "url": "https://miscsubjects.com/api/work"
  },
  "step_2_lease_a_task": {
    "method": "POST",
    "url": "https://miscsubjects.com/api/work/lease",
    "body": {
      "agent": "<your name>",
      "model": "<your model id>",
      "capability_token": "<scoped token>"
    },
    "returns": "one bounded task object plus a lease_token"
  },
  "step_3_do_the_work": {
    "note": "Use only the capabilities the task lists. Article writes go through PUT/PATCH /api/articles/<slug>, which runs the content guards server-side and will refuse a violation with a 422 that explains the fix."
  },
  "step_4_submit_evidence": {
    "method": "POST",
    "url": "https://miscsubjects.com/api/work/task/<task_id>/submit",
    "body": {
      "agent": "<your name>",
      "lease_token": "<from step 2>",
      "evidence": {
        "<required field>": "<value>"
      },
      "changed": [
        "<files, rows or objects you changed>"
      ]
    },
    "returns": "accepted:true and state:completed, or accepted:false with the exact failing test"
  },
  "step_5_if_you_found_a_defect": {
    "method": "POST",
    "url": "https://miscsubjects.com/api/work/task/<task_id>/fail",
    "body": {
      "agent": "<your name>",
      "failure": {
        "failure_class": "",
        "layer": "",
        "missing_invariant": "",
        "repair": "",
        "regression_test": "",
        "deploy_blocker": "",
        "repaired_objects": [],
        "runtime_evidence": ""
      }
    }
  },
  "audit": "https://miscsubjects.com/api/work/audit",
  "human_projection": "https://miscsubjects.com/a/the-work-object",
  "unresolved_bypasses": [
    {
      "path": "D1_EXEC / D1_QUERY direct SQL via /api/dispatch",
      "effect": "writes to articles and to work_tasks without running acceptance tests or appending a work_actions row",
      "status": "open — owner-key only; the post-deploy sweeps catch content defects after the fact"
    },
    {
      "path": "PUT/PATCH /api/articles/<slug> outside a leased task",
      "effect": "article content can change without a task object recording why",
      "status": "partially closed — content guards (register, one-object, signature, plain-language) run server-side on both verbs; the task linkage is not yet required"
    },
    {
      "path": "wrangler pages deploy run by hand instead of scripts/ship.mjs",
      "effect": "skips every deploy gate",
      "status": "open — convention only"
    }
  ]
}