Content
67%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.
A well-structured doctrine document whose layout rules are unusually concrete (exact center values, band percentages, embed spacing limits) and whose drop/rail/embed model is unambiguous. Its main weakness is redundancy — the overlay law and embed-scarcity rules are each restated multiple times, so the body could be roughly a third shorter with no loss of guidance.
Suggestions
Cut the "Why these two rules are one doctrine" section — it restates the overlay law and embed-scarcity rules already covered in the two core sections; keep at most its one-line framing in the intro.
State each load-bearing rule once: merge the duplicated embed warnings ("Embedding every word is the common mistake" vs "Embedding everything is the default mistake") into a single statement in the embed-scarce rules list.
Add minimal concreteness to the fuzzy directives: one example of an embed "designed entrance + exit" or what an inline emphasis highlight looks like, since these are the only instructions without parameters.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The overlay law is stated three times (intro paragraph, "The overlay law" section, and "Why these two rules are one doctrine"), and the embed-scarcity warning appears twice ("Embedding every word is the common mistake" / "Embedding everything is the default mistake"), so the closing section is largely redundant restatement. Nothing explains concepts Claude already knows, but the repetition is unnecessary padding that could be tightened into the two core sections. Not 2 because each section individually is dense with rule content rather than filler; not 4 because a whole section ("Why these two rules are one doctrine") could be cut with no loss of guidance. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is concrete: y = H/2 with exact values ("landscape 540, portrait 960"), "roughly the bottom ~5-8% of canvas height", "bottom ~80px center span", "≤1 embed per beat, never two adjacent/co-visible, spaced ≥ a beat apart", plus a clear drop/rail/embed decision table. Not 5 because some directives stay abstract — "designed entrance + exit" for embeds and "A punch word can get an inline emphasis highlight" give no concrete parameters or examples of what those look like; not 3 because the numeric layout rules and decision table are directly executable. | 4 / 5 |
Workflow Clarity | This is a single-purpose doctrine skill with no multi-step process, and the decision procedure is unambiguous: classify every phrase via the drop/rail/embed table, keep embeds scarce and spaced, compose at true center with no reserved band, and "Snapshot review judges caption-over-content legibility visually" names the verification step. Not 5 because the repeated restatement of the same rules in three places slightly obscures which formulation is load-bearing; not 3 because the operative rules ("Rail-first for talking-head / explainer", "Embed is scarce + spaced") are explicit and the visual-review checkpoint is stated. | 4 / 5 |
Progressive Disclosure | The skill is a single self-contained file with no bundle (no references/, scripts/, or assets/ exist), well organized into "The caption model", "The overlay law", and a synthesis section, and its one external pointer ("Applies ON TOP of the /hyperframes skill's references/captions.md") is clearly signaled and one level deep. Not 5 because the body runs ~73 lines with a redundant final section that should either be folded into the two core sections or trimmed, leaving the file tighter; not 3 because nothing that belongs in a separate file is inlined and navigation is easy. | 4 / 5 |
Total | 15 / 20 Passed |