Content
82%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a high-quality, expert-level skill body: nearly every paragraph carries non-obvious decision guidance, the carve workflow is fully executable with a real script and validation commands, and exact JSON is properly deferred to four real, well-signaled reference files. The main costs are length — the body is a full manual with some general-audio explanation and repetitive carve emphasis — and a workflow that must be assembled from prose sections rather than one explicit sequence.
Suggestions
Trim or push to a reference the general audio-engineering explanations (what compression, limiting, and gating do) and the repeated carve-rule emphasis — Claude already knows these basics, and consolidating the carve rules would shorten the body without losing framework-specific guidance.
Add one explicit ordered overview of the end-to-end mix workflow (diagnose -> choose preset/job -> order the chain -> carve if music under voice -> verify by render/listen) near the top, so the sequence and its checkpoints don't have to be assembled from prose across five sections.
Consider moving the detailed voiceover-carve operational material (group membership rules, strength behavior, lint rules) into its own reference file alongside presets.md, keeping SKILL.md as the overview with the command, the invariant, and the pointer.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-obvious, framework-specific knowledge (carve semantics, clip vs composition automation clocks, lint rules, render-refusal behavior) and avoids filler, but includes explanations of audio concepts Claude already knows — "Compression narrows the distance between loud and quiet", "A limiter is a ceiling" — plus some narrative repetition around the carve rules, fitting 'efficient; minor instances of over-explanation that could be trimmed' rather than the lean level-5. | 4 / 5 |
Actionability | Guidance is fully executable: the copy-paste command `node <SKILL_DIR>/scripts/carve.mjs --comp index.html` with documented flags (--bed/--voice/--strength/--dry-run) and sample output, a ready-to-use `<hf-audio-group>` HTML snippet, a symptom-to-preset table, and `npx hyperframes check` for confirmation — covering the common cases (carve, bus, preset fixes). | 5 / 5 |
Workflow Clarity | The mix workflow is sequenced with validation checkpoints (diagnose against in-file reference, symptom table, filter-then-level-then-character ordering, 'run the command... confirm... with npx hyperframes check, and only then render', and a listening feedback loop: 'If the bed sounds notched... the strength is too high'). However the sequence lives in prose across several sections rather than one explicit ordered process, and some validation is left implicit ('Read the ids back out of the chain rather than assuming'), matching level 4 rather than the fully checkpointed level 5. | 4 / 5 |
Progressive Disclosure | The four bundle references (attributes.md, fx-registry.md, diagnosis.md, presets.md) all exist, are one level deep, and are each clearly signaled with a purpose statement and read-order guidance ('read that before hand-building a chain'), which is level-5 behavior. But the body itself is a ~480-line manual whose extensive voiceover-carve operational detail could be its own reference file, and it points at paths outside the bundle (../hyperframes/references/plugin-installation.md, references/creator-editing-recipes.md, docs/reference/speed-ramps) — minor organization gaps that fit level 4. | 4 / 5 |
Total | 17 / 20 Passed |