{
  "eval_id": "judgmentkit-ui-generation-paired-artifact-v1",
  "evaluation_type": "live_provider_ui_generation_scoring",
  "generation_policy": "Calls a live provider for each baseline and JudgmentKit-guided variant, writes dated generated HTML artifacts, then scores those artifacts.",
  "visual_evidence": {
    "capture_engine": "chrome_devtools_protocol",
    "capture_policy": "Initial viewport screenshots captured from dated live-generated artifacts. Visual evidence is not used for scoring.",
    "viewports": [
      {
        "id": "desktop",
        "label": "Desktop",
        "width": 1365,
        "height": 900,
        "device_scale_factor": 1,
        "mobile": false
      },
      {
        "id": "mobile",
        "label": "Mobile",
        "width": 390,
        "height": 844,
        "device_scale_factor": 1,
        "mobile": true
      }
    ]
  },
  "benchmark_policy": "Qualitative paired-artifact evidence from live provider-generated artifacts only; not a statistically powered benchmark.",
  "claim_level": "repeated_pair_signal",
  "run": {
    "date": "2026-07-03",
    "mcp_release": "0.6.5",
    "mcp_release_segment": "mcp-0.6.5",
    "run_id": "run-002",
    "run_path": "2026-07-03/mcp-0.6.5/run-002",
    "html_report": "2026-07-03/mcp-0.6.5/run-002/ui-generation-report.html",
    "json_report": "2026-07-03/mcp-0.6.5/run-002/ui-generation-report.json"
  },
  "summary": {
    "cases": 2,
    "passed": 2,
    "failed": 0,
    "guided_wins": 2,
    "baseline_wins": 0,
    "ties": 0,
    "publishable": true,
    "publishability_status": "publishable",
    "publish_blockers": [],
    "publishability_policy": "Conservative UI eval publication requires no failed cases, baseline wins, or ties; fixture runs are never publishable as live evidence."
  },
  "metric_scale": {
    "metric_score": "0-5",
    "total_score": "0-100 weighted"
  },
  "generation": {
    "mode": "live_provider",
    "live": true,
    "provider": "codex",
    "model": "gpt-5.5",
    "model_config_id": "gpt-5-5-codex",
    "reasoning_effort": "high",
    "runtime": "codex exec",
    "local": false,
    "generated_artifacts_dir": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts",
    "artifacts": [
      {
        "case_id": "refund-triage-standalone-v1",
        "variant_id": "baseline",
        "treatment": "raw_brief_baseline",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:24:34.697Z",
        "duration_ms": 92784,
        "prompt_sha256": "952bfee6595f24d3f8a5c52a1e8f6cd40e18fc3d5b9640def71881ba38ae3423",
        "raw_response_sha256": "d7958778d06eddc582aa0d3cad83dc982ac1f56b640563716c48bd7282cb21ac"
      },
      {
        "case_id": "refund-triage-standalone-v1",
        "variant_id": "guided",
        "treatment": "judgmentkit_handoff",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:26:07.483Z",
        "duration_ms": 139533,
        "prompt_sha256": "ee886720a0fc0960b038566e8379289eea789cf403e2033797390d3ae06b12b9",
        "raw_response_sha256": "471e02e9678f3da431a31f4aa7e6a50f5e73e6c69f98b8cf7bc3629a3eccb0eb"
      },
      {
        "case_id": "dinner-playlist-standalone-v1",
        "variant_id": "baseline",
        "treatment": "raw_brief_baseline",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:28:27.031Z",
        "duration_ms": 433457,
        "prompt_sha256": "2517f2809485e70391260a9242552d4257e18736db836fd2d4600f79ecd8d523",
        "raw_response_sha256": "ebf860333e9d831817a42d0cc9c02b623475b0da29fcf3e23a3af29458a23d9f"
      },
      {
        "case_id": "dinner-playlist-standalone-v1",
        "variant_id": "guided",
        "treatment": "judgmentkit_handoff",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:35:40.489Z",
        "duration_ms": 139675,
        "prompt_sha256": "1f911ff177f0628bb06abeb7cfbd626dc6f0c9cdd7dc8666d68e9890603db9d1",
        "raw_response_sha256": "76645d7620e145b971571b0c3c43b65de05d2b3b80c51ff40a7c407b5e09ac40"
      }
    ]
  },
  "live_generation": {
    "policy": "Calls a live provider for each baseline and JudgmentKit-guided variant, writes dated generated HTML artifacts, then scores those artifacts.",
    "provider_metadata": {
      "provider": "codex",
      "model_config_id": "gpt-5-5-codex",
      "model": "gpt-5.5",
      "runtime": "codex exec",
      "local": false
    },
    "generated_artifacts_dir": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts",
    "artifacts": [
      {
        "case_id": "refund-triage-standalone-v1",
        "variant_id": "baseline",
        "treatment": "raw_brief_baseline",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:24:34.697Z",
        "duration_ms": 92784,
        "prompt_sha256": "952bfee6595f24d3f8a5c52a1e8f6cd40e18fc3d5b9640def71881ba38ae3423",
        "raw_response_sha256": "d7958778d06eddc582aa0d3cad83dc982ac1f56b640563716c48bd7282cb21ac"
      },
      {
        "case_id": "refund-triage-standalone-v1",
        "variant_id": "guided",
        "treatment": "judgmentkit_handoff",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:26:07.483Z",
        "duration_ms": 139533,
        "prompt_sha256": "ee886720a0fc0960b038566e8379289eea789cf403e2033797390d3ae06b12b9",
        "raw_response_sha256": "471e02e9678f3da431a31f4aa7e6a50f5e73e6c69f98b8cf7bc3629a3eccb0eb"
      },
      {
        "case_id": "dinner-playlist-standalone-v1",
        "variant_id": "baseline",
        "treatment": "raw_brief_baseline",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:28:27.031Z",
        "duration_ms": 433457,
        "prompt_sha256": "2517f2809485e70391260a9242552d4257e18736db836fd2d4600f79ecd8d523",
        "raw_response_sha256": "ebf860333e9d831817a42d0cc9c02b623475b0da29fcf3e23a3af29458a23d9f"
      },
      {
        "case_id": "dinner-playlist-standalone-v1",
        "variant_id": "guided",
        "treatment": "judgmentkit_handoff",
        "artifact": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
        "repo_artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
        "provider": "codex",
        "model": "gpt-5.5",
        "generated_at": "2026-07-03T17:35:40.489Z",
        "duration_ms": 139675,
        "prompt_sha256": "1f911ff177f0628bb06abeb7cfbd626dc6f0c9cdd7dc8666d68e9890603db9d1",
        "raw_response_sha256": "76645d7620e145b971571b0c3c43b65de05d2b3b80c51ff40a7c407b5e09ac40"
      }
    ]
  },
  "provider_metadata": {
    "provider": "codex",
    "model_config_id": "gpt-5-5-codex",
    "model": "gpt-5.5",
    "runtime": "codex exec",
    "local": false
  },
  "results": [
    {
      "id": "refund-triage-standalone-v1",
      "title": "Refund triage handoff",
      "task_prompt": "Review the selected refund request and prepare the next handoff.",
      "claim_level": "single_pair_signal",
      "expected_outcomes": [
        "The reviewer can identify the selected case.",
        "The reviewer can choose approve, policy review, or return for evidence.",
        "The reviewer can complete a handoff with a reason and next owner."
      ],
      "winner": "judgmentkit_handoff",
      "expected_winner": "judgmentkit_handoff",
      "score_delta": 36.08,
      "minimum_score_delta": 20,
      "passed": true,
      "variants": [
        {
          "id": "baseline",
          "label": "Version A",
          "treatment": "raw_brief_baseline",
          "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
          "public_artifact": null,
          "live_generation": {
            "mode": "live_provider",
            "provider": "codex",
            "model": "gpt-5.5",
            "model_config_id": "gpt-5-5-codex",
            "reasoning_effort": "high",
            "generated_at": "2026-07-03T17:24:34.697Z",
            "duration_ms": 92784,
            "prompt_sha256": "952bfee6595f24d3f8a5c52a1e8f6cd40e18fc3d5b9640def71881ba38ae3423",
            "raw_response_sha256": "d7958778d06eddc582aa0d3cad83dc982ac1f56b640563716c48bd7282cb21ac",
            "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
            "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
            "generation_notes": [
              "Built as a standalone support operations triage surface for case R-1842.",
              "Included the implementation-heavy fields requested in the source brief while keeping the primary workflow centered on decision and handoff.",
              "Preselected return for evidence because the receipt photo is missing and manager approval requires complete evidence."
            ],
            "execution": {
              "runtime": "codex exec",
              "status": 0,
              "stdout_sha256": "5908c996704d14fafd0c5591e9b27ea50b5a9dc180a4ea683c5235a28fb5276a",
              "stderr_sha256": "6edd6098b52e55c18f74995e84a969c4b187794527554fd38907205983c7bf58",
              "command_display": "codex exec --model gpt-5.5 -c model_reasoning_effort=\"high\" --sandbox read-only --skip-git-repo-check --ephemeral --color never --output-schema <schema> --output-last-message <output> -"
            }
          },
          "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
          "generated_artifact": {
            "path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html",
            "repo_path": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html"
          },
          "metadata_treatment": "raw_brief_baseline",
          "metadata_comparison_id": "refund-triage-standalone-v1",
          "score": 55.92,
          "metric_results": {
            "activity_fit": {
              "score": 1,
              "present": [
                "Evidence checklist"
              ],
              "missing": [
                "Daily triage",
                "Refund escalation queue",
                "Customer refund escalation",
                "Policy review context"
              ]
            },
            "decision_support": {
              "score": 3.75,
              "present": [
                "Approve refund",
                "Send to policy review",
                "Return for evidence"
              ],
              "missing": [
                "Choose a path"
              ]
            },
            "disclosure_discipline": {
              "score": 4.25,
              "implementation_leakage": [
                "JSON schema"
              ],
              "review_packet_leakage": [],
              "leakage_count": 1
            },
            "handoff_completeness": {
              "score": 3,
              "present": [
                "Handoff",
                "Next owner",
                "Support agent"
              ],
              "missing": [
                "Policy reviewer",
                "Handoff reason"
              ]
            },
            "task_success_support": {
              "score": 1,
              "present": [
                "Receipt photo is missing"
              ],
              "missing": [
                "Review selected case",
                "Check evidence",
                "Choose refund path",
                "Prepare handoff"
              ]
            },
            "confidence_rework_signals": {
              "score": 1.67,
              "present": [
                "Evidence checklist"
              ],
              "missing": [
                "Policy review context",
                "missing receipt photo"
              ]
            }
          },
          "screenshots": [
            {
              "id": "baseline-desktop",
              "label": "Version A Desktop",
              "viewport": {
                "id": "desktop",
                "label": "Desktop",
                "width": 1365,
                "height": 900,
                "device_scale_factor": 1,
                "mobile": false
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/refund-triage-standalone-v1/baseline-desktop.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html"
            },
            {
              "id": "baseline-mobile",
              "label": "Version A Mobile",
              "viewport": {
                "id": "mobile",
                "label": "Mobile",
                "width": 390,
                "height": 844,
                "device_scale_factor": 1,
                "mobile": true
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/refund-triage-standalone-v1/baseline-mobile.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/baseline.html"
            }
          ]
        },
        {
          "id": "guided",
          "label": "Version B",
          "treatment": "judgmentkit_handoff",
          "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
          "public_artifact": null,
          "live_generation": {
            "mode": "live_provider",
            "provider": "codex",
            "model": "gpt-5.5",
            "model_config_id": "gpt-5-5-codex",
            "reasoning_effort": "high",
            "generated_at": "2026-07-03T17:26:07.483Z",
            "duration_ms": 139533,
            "prompt_sha256": "ee886720a0fc0960b038566e8379289eea789cf403e2033797390d3ae06b12b9",
            "raw_response_sha256": "471e02e9678f3da431a31f4aa7e6a50f5e73e6c69f98b8cf7bc3629a3eccb0eb",
            "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
            "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
            "generation_notes": [
              "Standalone artifact uses only semantic HTML and scoped CSS.",
              "Primary surface translates the source request into refund triage language and avoids implementation details in the visible UI.",
              "The workspace keeps queue context, selected case evidence, policy context, decision choice, and completion handoff together."
            ],
            "execution": {
              "runtime": "codex exec",
              "status": 0,
              "stdout_sha256": "6da2519b7b51efc9b9df9eb7c298f50d7d10fa60d12b61292f10cde1c2ec8b63",
              "stderr_sha256": "f6599b92c86d586b5e98d314305ac2f6db6a39ed4c3cd60dcf4554bf63bf07e9",
              "command_display": "codex exec --model gpt-5.5 -c model_reasoning_effort=\"high\" --sandbox read-only --skip-git-repo-check --ephemeral --color never --output-schema <schema> --output-last-message <output> -"
            }
          },
          "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
          "generated_artifact": {
            "path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html",
            "repo_path": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html"
          },
          "metadata_treatment": "judgmentkit_handoff",
          "metadata_comparison_id": "refund-triage-standalone-v1",
          "score": 92,
          "metric_results": {
            "activity_fit": {
              "score": 5,
              "present": [
                "Daily triage",
                "Refund escalation queue",
                "Customer refund escalation",
                "Evidence checklist",
                "Policy review context"
              ],
              "missing": []
            },
            "decision_support": {
              "score": 5,
              "present": [
                "Choose a path",
                "Approve refund",
                "Send to policy review",
                "Return for evidence"
              ],
              "missing": []
            },
            "disclosure_discipline": {
              "score": 5,
              "implementation_leakage": [],
              "review_packet_leakage": [],
              "leakage_count": 0
            },
            "handoff_completeness": {
              "score": 4,
              "present": [
                "Handoff",
                "Next owner",
                "Support agent",
                "Policy reviewer"
              ],
              "missing": [
                "Handoff reason"
              ]
            },
            "task_success_support": {
              "score": 3,
              "present": [
                "Review selected case",
                "Choose refund path",
                "Receipt photo is missing"
              ],
              "missing": [
                "Check evidence",
                "Prepare handoff"
              ]
            },
            "confidence_rework_signals": {
              "score": 5,
              "present": [
                "Policy review context",
                "Evidence checklist",
                "missing receipt photo"
              ],
              "missing": []
            }
          },
          "screenshots": [
            {
              "id": "guided-desktop",
              "label": "Version B Desktop",
              "viewport": {
                "id": "desktop",
                "label": "Desktop",
                "width": 1365,
                "height": 900,
                "device_scale_factor": 1,
                "mobile": false
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/refund-triage-standalone-v1/guided-desktop.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html"
            },
            {
              "id": "guided-mobile",
              "label": "Version B Mobile",
              "viewport": {
                "id": "mobile",
                "label": "Mobile",
                "width": 390,
                "height": 844,
                "device_scale_factor": 1,
                "mobile": true
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/refund-triage-standalone-v1/guided-mobile.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/refund-triage-standalone-v1/guided.html"
            }
          ]
        }
      ],
      "rationale": [
        "JudgmentKit-guided artifact scored 36.08 points above baseline.",
        "Implementation leakage changed from 1 baseline terms to 0 guided terms.",
        "Activity-fit evidence changed from 1 matched terms to 5 matched terms."
      ]
    },
    {
      "id": "dinner-playlist-standalone-v1",
      "title": "Dinner playlist builder",
      "task_prompt": "Build a 10-song dinner playlist that starts mellow, lifts in the middle, avoids disliked artists and explicit tracks, and leaves a sequence note.",
      "claim_level": "single_pair_signal",
      "expected_outcomes": [
        "The host can assemble a 10-song playlist.",
        "The host can catch explicit and disliked-artist conflicts.",
        "The host can explain the sequence from mellow start through lifted middle to soft close."
      ],
      "winner": "judgmentkit_handoff",
      "expected_winner": "judgmentkit_handoff",
      "score_delta": 41.65,
      "minimum_score_delta": 20,
      "passed": true,
      "variants": [
        {
          "id": "baseline",
          "label": "Version A",
          "treatment": "raw_brief_baseline",
          "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
          "public_artifact": null,
          "live_generation": {
            "mode": "live_provider",
            "provider": "codex",
            "model": "gpt-5.5",
            "model_config_id": "gpt-5-5-codex",
            "reasoning_effort": "high",
            "generated_at": "2026-07-03T17:28:27.031Z",
            "duration_ms": 433457,
            "prompt_sha256": "2517f2809485e70391260a9242552d4257e18736db836fd2d4600f79ecd8d523",
            "raw_response_sha256": "ebf860333e9d831817a42d0cc9c02b623475b0da29fcf3e23a3af29458a23d9f",
            "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
            "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
            "generation_notes": [
              "Built as a standalone product surface with no external assets, scripts, imports, or hidden benchmark terms in the HTML.",
              "Includes the source-requested implementation details while keeping the primary workflow focused on playlist selection, ordering, conflicts, save/share, and sequence note.",
              "Uses semantic controls for playlist actions, conflict marking, filters, table review, and note editing."
            ],
            "execution": {
              "runtime": "codex exec",
              "status": 0,
              "stdout_sha256": "309118b11245a0894a2b5d3435cd23ef75abcd4148f43fb66823d2f558d0f105",
              "stderr_sha256": "b192f60f08662ea9a49a42424fc884ace2d48ae3000a29b2ec7ba8d4c11f59db",
              "command_display": "codex exec --model gpt-5.5 -c model_reasoning_effort=\"high\" --sandbox read-only --skip-git-repo-check --ephemeral --color never --output-schema <schema> --output-last-message <output> -"
            }
          },
          "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
          "generated_artifact": {
            "path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html",
            "repo_path": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html"
          },
          "metadata_treatment": "raw_brief_baseline",
          "metadata_comparison_id": "dinner-playlist-standalone-v1",
          "score": 36.82,
          "metric_results": {
            "activity_fit": {
              "score": 1.67,
              "present": [
                "Guest preferences",
                "Sequence note"
              ],
              "missing": [
                "Dinner brief",
                "Suggested tracks",
                "Playlist sequence",
                "Conflict checks"
              ]
            },
            "decision_support": {
              "score": 2,
              "present": [
                "Add to playlist",
                "Mark as conflict"
              ],
              "missing": [
                "Move earlier",
                "Move later",
                "Remove track"
              ]
            },
            "disclosure_discipline": {
              "score": 0,
              "implementation_leakage": [
                "data model",
                "track table field",
                "JSON schema",
                "prompt template",
                "tool call",
                "resource id",
                "API endpoint",
                "CRUD",
                "field",
                "model"
              ],
              "review_packet_leakage": [
                "candidate"
              ],
              "leakage_count": 11
            },
            "handoff_completeness": {
              "score": 3.75,
              "present": [
                "Save playlist",
                "Share playlist",
                "Sequence note"
              ],
              "missing": [
                "10 tracks ready to play"
              ]
            },
            "task_success_support": {
              "score": 3.57,
              "present": [
                "explicit track",
                "disliked artist",
                "mellow opener",
                "warm middle",
                "closing track"
              ],
              "missing": [
                "genre balance",
                "energy flow"
              ]
            },
            "confidence_rework_signals": {
              "score": 0,
              "present": [],
              "missing": [
                "Fits brief",
                "Conflict checks",
                "ready to play"
              ]
            }
          },
          "screenshots": [
            {
              "id": "baseline-desktop",
              "label": "Version A Desktop",
              "viewport": {
                "id": "desktop",
                "label": "Desktop",
                "width": 1365,
                "height": 900,
                "device_scale_factor": 1,
                "mobile": false
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/dinner-playlist-standalone-v1/baseline-desktop.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html"
            },
            {
              "id": "baseline-mobile",
              "label": "Version A Mobile",
              "viewport": {
                "id": "mobile",
                "label": "Mobile",
                "width": 390,
                "height": 844,
                "device_scale_factor": 1,
                "mobile": true
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/dinner-playlist-standalone-v1/baseline-mobile.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/baseline.html"
            }
          ]
        },
        {
          "id": "guided",
          "label": "Version B",
          "treatment": "judgmentkit_handoff",
          "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
          "public_artifact": null,
          "live_generation": {
            "mode": "live_provider",
            "provider": "codex",
            "model": "gpt-5.5",
            "model_config_id": "gpt-5-5-codex",
            "reasoning_effort": "high",
            "generated_at": "2026-07-03T17:35:40.489Z",
            "duration_ms": 139675,
            "prompt_sha256": "1f911ff177f0628bb06abeb7cfbd626dc6f0c9cdd7dc8666d68e9890603db9d1",
            "raw_response_sha256": "76645d7620e145b971571b0c3c43b65de05d2b3b80c51ff40a7c407b5e09ac40",
            "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
            "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
            "generation_notes": [
              "Domain language is used for playlist decisions, guest preferences, conflicts, ordering, and handoff actions.",
              "The product surface keeps implementation details out of the user-facing HTML and presents conflicts as dinner-playlist evidence.",
              "Responsive CSS keeps the workspace usable as one column on narrow screens, with visible focus states and 44px primary controls."
            ],
            "execution": {
              "runtime": "codex exec",
              "status": 0,
              "stdout_sha256": "a1af1483f030a282a5449f6bfa7770eaab7e6816db35e7aad9d83dba8ee37218",
              "stderr_sha256": "ca4d41380439fa05f583739bc0cce6e829f221170e215f51b152e66b9fbff50a",
              "command_display": "codex exec --model gpt-5.5 -c model_reasoning_effort=\"high\" --sandbox read-only --skip-git-repo-check --ephemeral --color never --output-schema <schema> --output-last-message <output> -"
            }
          },
          "generated_artifact_path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
          "generated_artifact": {
            "path": "2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html",
            "repo_path": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html"
          },
          "metadata_treatment": "judgmentkit_handoff",
          "metadata_comparison_id": "dinner-playlist-standalone-v1",
          "score": 78.47,
          "metric_results": {
            "activity_fit": {
              "score": 5,
              "present": [
                "Dinner brief",
                "Guest preferences",
                "Suggested tracks",
                "Playlist sequence",
                "Conflict checks",
                "Sequence note"
              ],
              "missing": []
            },
            "decision_support": {
              "score": 2,
              "present": [
                "Add to playlist",
                "Mark as conflict"
              ],
              "missing": [
                "Move earlier",
                "Move later",
                "Remove track"
              ]
            },
            "disclosure_discipline": {
              "score": 5,
              "implementation_leakage": [],
              "review_packet_leakage": [],
              "leakage_count": 0
            },
            "handoff_completeness": {
              "score": 3.75,
              "present": [
                "Save playlist",
                "Share playlist",
                "Sequence note"
              ],
              "missing": [
                "10 tracks ready to play"
              ]
            },
            "task_success_support": {
              "score": 3.57,
              "present": [
                "explicit track",
                "disliked artist",
                "mellow opener",
                "warm middle",
                "closing track"
              ],
              "missing": [
                "genre balance",
                "energy flow"
              ]
            },
            "confidence_rework_signals": {
              "score": 3.33,
              "present": [
                "Conflict checks",
                "ready to play"
              ],
              "missing": [
                "Fits brief"
              ]
            }
          },
          "screenshots": [
            {
              "id": "guided-desktop",
              "label": "Version B Desktop",
              "viewport": {
                "id": "desktop",
                "label": "Desktop",
                "width": 1365,
                "height": 900,
                "device_scale_factor": 1,
                "mobile": false
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/dinner-playlist-standalone-v1/guided-desktop.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html"
            },
            {
              "id": "guided-mobile",
              "label": "Version B Mobile",
              "viewport": {
                "id": "mobile",
                "label": "Mobile",
                "width": 390,
                "height": 844,
                "device_scale_factor": 1,
                "mobile": true
              },
              "path": "2026-07-03/mcp-0.6.5/run-002/screenshots/dinner-playlist-standalone-v1/guided-mobile.png",
              "artifact": "evals/reports/2026-07-03/mcp-0.6.5/run-002/generated-artifacts/dinner-playlist-standalone-v1/guided.html"
            }
          ]
        }
      ],
      "rationale": [
        "JudgmentKit-guided artifact scored 41.65 points above baseline.",
        "Implementation leakage changed from 10 baseline terms to 0 guided terms.",
        "Activity-fit evidence changed from 2 matched terms to 6 matched terms."
      ]
    }
  ]
}