Results

Every tool returns both prose and structuredContent:

json
{
  "outcome": "conformant",
  "passed": 5, "failed": 0, "errored": 0, "skipped": 1,
  "results": [
    { "id": "MPP-11", "name": "...", "status": "PASS",
      "detail": "...", "destructive": false }
  ]
}

outcome is a name, not an exit code — an integer means nothing to an agent, and no-verdict must never be read as conformant. When any check errored, the prose channel carries an explicit caveat that no statement was made about the target's conformance.