Conciseness

The body is dense with operational detail and zero padding: no concept explanations, no library intros, every sentence carries a flag, default, or constraint (e.g. "plate 25% above / 75% below", "cap ~0.019 H", "~14 characters a line"). It matches the score-5 anchor — lean, assumes Claude's competence, every token earns its place — and is well above the score-4 'minor instances of over-explanation' case.

Actionability

Every workflow is given as a copy-paste-ready command with real flags and defaults: `python captions.py --video reel.mp4 --beats cutlist.aligned.json --words reel.words.json --out final.mp4 [--style plate|outline]`, plus concrete card syntax `--card "LINE ONE|LINE TWO" --card-until 4.7`. The beats input format is specified ("Any file with `beats: [{start, end, vo}]` works"), covering the common cases like the description's examples do at score 5.

Workflow Clarity

The sequence (transcribe finished audio → captions burn → inspect) is explicit, and the batch render operation has real validation checkpoints and feedback loops: "regenerate after alignment or copy changes", "Preview rejects a changed cut list until its footprint is rebuilt", "Inspect final captioned frames as well as this planned coverage", and "An explicitly supplied missing font fails in both commands" documents error behavior. This matches the score-5 anchor (explicit validation steps plus error-recovery loops) for a batch operation that would otherwise be capped at 3.

Progressive Disclosure

The body is well organized into Run / Placement and style / plates.py / Rules / Footprint sections and the scripts it names (transcribe.py, captions.py, plates.py) exist in the scripts/ bundle with commands shown. However the footprint tool is referred to only as "the bundled footprint helper" without naming footprint.py or giving its command line, and media_proxy.py/_fonts.py are never mentioned — minor organization gaps that fit the score-4 anchor rather than the fully clear score-5 navigation.

Specificity

The description lists multiple specific concrete actions with comprehensive coverage: "gets word timings from the video's own audio", "burns one to three words at a time with Pillow + ffmpeg (no libass needed)", "pinned to a split-screen seam (plate 25% above / 75% below) or at a fixed height", "plate, outline or one-word serif style", "optional red hook card", and "burns per-beat caption blocks (black, one union silhouette, placed in the emptiest band)". It matches the score-5 anchor of multiple specific concrete actions covering all three caption kinds plus the transcription step.

Completeness

It clearly answers what (three caption kinds, which script does what, styles, placement, CTA hold behavior) and when via an explicit trigger clause: "Use as the last step of any video ad." The 'when' is explicit and concrete rather than weakly implied, matching the score-5 anchor; without that clause completeness would have been capped at 3.

Trigger Term Quality

Natural terms a user would say are present — "captions", "burned-in captions", "vertical video", "video ad", "transcribe", "no voice" — but common variations are missing: no "subtitles" or "CC" synonyms, and no file extension (".mp4", ".mov"). This fits the score-4 anchor (good keyword coverage, a few natural terms missing) rather than 5 (comprehensive synonyms and extensions) or 3 (missing common variations broadly).

Distinctiveness Conflict Risk

It occupies a clear niche — burned-in captions for finished vertical video ads via named scripts (transcribe.py/captions.py/plates.py) with a GooseWorks/fal Whisper proxy and per-beat caption blocks — with triggers unlikely to fire for unrelated skills. It matches the score-5 anchor (clear niche, distinct triggers, minimal conflict risk); it is far more specific than the score-4 'Works with PDF and Word document files' example.