Conciseness

The body is lean and operational throughout ("renders never fetch", "No signal, no suggestion; respect 'leave it'", "--plan is for you"), adding only project-specific policy Claude could not know, with zero padding or explanation of known concepts. Not 4: there are no sections that could be trimmed without losing load-bearing guidance.

Actionability

The core verb is copy-paste ready with an output contract ("npx hyperframes media-use resolve --type <type> --intent …"; "prints one line, resolved <id> → <path>"), the table supplies concrete --type values (bgm, sfx, logo, voice, video, cutout), and exact commands (auth status, tts, transcribe, remove-background, media-treatment) plus a credit-limit JSON snippet cover the common cases. Not 4: no missing key details for the primary flow.

Workflow Clarity

Checkpoints and feedback loops are present (auth status before paid calls, wait until the person confirms sign-in; a refused call names a billing command - say why, ask them to run it, then retry; --candidates before generating), fitting the clear-sequence-with-most-checkpoints anchor. It is not 5 because the steps are distributed across sections rather than consolidated into one explicit ordered flow, and not 3 because validation for the risky paths (auth, billing) is explicit.

Progressive Disclosure

Scored against the actual bundle: the "Where to look" table cleanly routes 10 tasks to 10 real reference files, but media-treatments.md points onward to media-treatment-recipes.md (a 2-level chain), meta.md and telemetry-dashboard.md are never linked from the body, and the 14-file scripts/ bundle is only surfaced indirectly through references. This fits good-structure-with-minor-organization-gaps rather than the clean one-level-deep anchor.

Specificity

The description enumerates concrete actions across the full media domain ("Get or make any media … music, sound effects, images, icons, logos, voiceover, avatar and generated video, transcripts, cutouts, colour grades and looks - with HeyGen or free local tools, and mix placed audio") with named providers, matching the comprehensive-coverage anchor. It is not 4 because the action set (get, make, mix, grade/treat, transcribe, cut out) leaves only trivial gaps.

Completeness

Both a clear 'what' (get/make media and mix placed audio with HeyGen or free local tools) and an explicit 'when' branch ("Also for footage that looks dark, flat or dull, or should be stylized or hidden") are present, so the missing-trigger cap at 3 does not apply. It is not 5 because the primary use case lacks an explicit "Use when…" clause - the trigger is only implied by "for a HyperFrames video".

Trigger Term Quality

Natural terms users would say are well covered ("music", "sound effects", "voiceover", "logos", "footage that looks dark, flat or dull", "stylized or hidden", "mix"), fitting the good-coverage anchor. It is not 5 because common synonyms and variations are absent (no "voice-over"/"VO", "SFX", "captions"/"subtitles", "B-roll").

Distinctiveness Conflict Risk

"for a HyperFrames video" scopes the skill to a clear niche with distinct triggers (media types, look/treatment feedback), giving minimal conflict risk with generic media or image skills. Not 4: no meaningful overlap risk remains within its ecosystem.