{
  "contract": "this hash is committed to git BEFORE the benchmark runs; a matching observed outcome then proves the prediction was pre-registered, not fit post-hoc.",
  "name": "s08_capinfer_estate_servers",
  "note": "S08 O11. Sealed before any estate MCP server was read or run.",
  "prediction": {
    "bar": {
      "T1_static_census_whole_list": "every MCP server in the nine trees is censused for whether it declares tool annotations at all (readOnlyHint / destructiveHint / idempotentHint / openWorldHint). Static read, no execution. A server that declares NO hints cannot be graded by capinfer at all, and that is the finding for that server.",
      "T2_observation_where_runnable": "full capinfer observation is attempted on f3's own oss/minicheck-mcp. Prediction sealed from the tools/list capture BEFORE observation, per the S06-1 protocol.",
      "T3_control_first": "the six pinned third-party servers must reproduce 15/15 in the same session before any estate server is graded (the S06-1 control).",
      "T4_kill_per_server": "a server that cannot be installed in capinfer-sandbox:1 without editing the pinned image, or that is an HTTP client of a live backend, is recorded NOT RUN with the reason and the cost. Not run is not a failure and is not a pass.",
      "no_hint_no_grade": "if a server declares no hints, capinfer has nothing to contradict. Reporting 'no disagreements' for such a server would be the estate's signature defect -- a green that means nothing. Those servers are reported UNGRADABLE, not clean."
    },
    "both_outcomes_publishable": true,
    "candidate": "S08-O11 capinfer over the estate's own MCP servers",
    "instrument": "capinfer (declared capability hints vs observed syscalls) plus a new static declaration census",
    "prediction_recorded_now": "most estate MCP servers declare no tool annotations at all, because the MCP annotation fields are optional and rarely set outside the reference servers. If so the estate's own servers are ungradable by capinfer today, and the actionable result is a one-line ask to each lane rather than a disagreement count."
  },
  "schema": "prereg-seal-v1",
  "sealed_at_commit": "54f7fdb68f3201633b7226a516f9634c44c3ac98",
  "sha256": "ec8cf38c0958b084ce4f9a795b01828a86ff451ae60e7bdf6d5a68d387289686"
}