Conciseness

The body is ~37 lean lines: format signature, the run commands, the non-obvious frame-swap rationale (Chromium can't decode H.264 over file://, hangs on canplay), and a contract list. Every fact is something Claude could not infer — nothing explains known concepts, no padding. This matches anchor 5 ('Lean and efficient; every token earns its place'); even the codec explanation is genuinely non-obvious and load-bearing.

Actionability

Concrete runnable commands are given with full flags and outputs ("build_card.py --config config.json --out hyperframe.html ; render.py --config config.json --html hyperframe.html --out master-silent.mp4") and both scripts exist in the bundle, matching anchor 4 ('mostly executable guidance... minor gaps'). Not 5 because the invocations omit interpreter/path context (python scripts/build_card.py) and no sample config or field list is shown, leaving the config schema implicit.

Workflow Clarity

The build-then-render sequence is present and ordered, plus the input constraint 'Provide clips longer than the master (or the swap just loops them)', but there are no validation checkpoints on a batch-style render (~10s of frame-stepped output). Per the guideline capping batch operations without validation at 3, this sits at anchor 3 ('sequence present but checkpoints missing or implicit') — the sequence itself is coherent, so not 2.

Progressive Disclosure

The ~37-line body is appropriately self-contained with well-marked sections (signature, ## Run, ## Contract) and correctly leaves tile/music generation to sibling capabilities — the under-50-lines exception applies. Scored 4 rather than 5 because the scripts build_card.py, render.py (and _shared.js) are referenced by bare filename without signaling they live in scripts/, a minor organization gap matching anchor 4.

Specificity

The description enumerates multiple concrete actions and outputs: "Render a 'photo-grid promo card' video from a config", "a real-DOM card with a FIXED header/footer and a CONTINUOUSLY SCROLLING 2-row grid", "frame-stepped via Playwright", "encoded with FFmpeg", covering tile types, tooling, and cost model comprehensively. This matches the anchor 'Lists multiple specific concrete actions; comprehensive coverage' — not 4, since there are no meaningful coverage gaps in what the capability does.

Completeness

The 'what' is thoroughly explicit (render, scroll, frame-step, encode) and a 'when' clause exists ("Use for the photo-grid-promo-card format"). This matches anchor 4 ('Has both what and when; when could be more explicit or specific') — the trigger is a single narrow format name rather than the concrete multi-condition triggers of anchor 5.

Trigger Term Quality

Relevant keywords exist ("photo-grid promo card", "promo-code", "CTA", "video", "config") but the only real trigger phrase is "Use for the photo-grid-promo-card format", a single format slug. This matches anchor 3 ('Some relevant keywords but missing common variations or synonyms') — it lacks natural variations a user might say ("promo video", "product ad", "grid video", "scrolling card"), so not 4.

Distinctiveness Conflict Risk

A highly specific niche format name plus explicit boundaries against sibling capabilities ("the clips + music come from create-video-fal + create-music-elevenlabs", "no paid calls in this capability") gives clear distinct triggers with minimal conflict risk, matching anchor 5. It is more self-delineating than anchor 4's 'minor overlap risk with closely related skills'.