docs/REVIEW-CASES.json

{
  "executionScope": "Synthetic local Node HTTP, local workerd, and production Cloudflare HTTPS tool tests. Exactly five positive and three negative service cases passed hosted. Assistant routing in ChatGPT/Codex is NOT tested. No account required for noauth.",
  "positive": [
    {
      "id": "P1",
      "scenario": "Browse templates",
      "prompt": "Show me the public capability templates.",
      "expectedTools": [
        "list_capability_templates"
      ],
      "expectedResult": "Three public templates, no private catalog.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    },
    {
      "id": "P2",
      "scenario": "Get a template",
      "prompt": "Show the recurring checklist template.",
      "expectedTools": [
        "get_capability_template"
      ],
      "expectedResult": "checklist template returned.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    },
    {
      "id": "P3",
      "scenario": "Validate a draft",
      "prompt": "Check this complete summary capability draft.",
      "expectedTools": [
        "validate_capability_draft"
      ],
      "expectedResult": "structural_pass true, needs_review true, not_runtime_tested true.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    },
    {
      "id": "P4",
      "scenario": "Render a draft",
      "prompt": "Render this complete draft as reviewable Markdown.",
      "expectedTools": [
        "render_capability_draft"
      ],
      "expectedResult": "Verbatim goal field preserved; escaped Markdown with stable claim IDs.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    },
    {
      "id": "P5",
      "scenario": "Revise safely",
      "prompt": "Revise my claim statement, retain its ID, and render this text containing HTML and a Markdown link.",
      "expectedTools": [
        "render_capability_draft"
      ],
      "expectedResult": "Returned output retains claim ID; HTML/link punctuation encoded as inert text.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    }
  ],
  "negative": [
    {
      "id": "N1",
      "scenario": "Missing goal",
      "prompt": "Render this draft without a goal.",
      "expectedTools": [
        "render_capability_draft"
      ],
      "expectedResult": "INVALID_INPUT; no rendered artifact.",
      "reason": "Required verbatim goal absent.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    },
    {
      "id": "N2",
      "scenario": "Unlinked evidence",
      "prompt": "Render this draft whose evidence references C-unknown.",
      "expectedTools": [
        "render_capability_draft"
      ],
      "expectedResult": "INVALID_DRAFT; no rendered artifact.",
      "reason": "Evidence must reference a stable existing claim.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    },
    {
      "id": "N3",
      "scenario": "Unsupported deployment",
      "prompt": "Deploy this capability now.",
      "expectedTools": [
        "deploy_capability (synthetic unsupported-operation probe)"
      ],
      "expectedResult": "JSON-RPC -32602 Invalid params; skill explains separate workflow needed.",
      "reason": "No deployment operation exists.",
      "localToolStatus": "passed",
      "hostReviewStatus": "not_run",
      "workerToolStatus": "passed_local_workerd",
      "hostedToolStatus": "passed_production_https"
    }
  ]
}