Content
93%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 lean, highly actionable body: copy-paste commands for search, install, variables, beats, and gap reporting, with details correctly delegated to real, well-labeled one-level-deep reference files. The only meaningful gap is that the find->install->wire sequence is conveyed by section order rather than explicit steps, and batch installs (`add <tag>`) lack a stated post-install verification checkpoint beyond the printed file list.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with operational facts and zero conceptual padding: "npx hyperframes catalog --query ... --json", "add needs the network even for an item installed before; search works offline from the cached manifest", "Keep an installed item's ids prefixed as they are, so two items can't collide". It never explains concepts Claude already knows (what HTML, compositions, or timelines are) and every line carries a rule, command, or pointer. It matches anchor 5 ('lean and efficient; assumes Claude's competence; every token earns its place') rather than 4, which reserves room for trimmable over-explanation — the only candidate, "Starting from an item is faster and usually better", is a one-sentence orientation that sets decision policy (prefer installed items over hand-building), not padding. | 5 / 5 |
Actionability | Every workflow step ships a copy-paste command: `npx hyperframes catalog --query "..." --json`, `npx hyperframes add <name> --json` ("prints the written files and the mount snippet"), `add --vars '{"title":"…"}'`, `npx hyperframes beats --json`, and a pre-filled gap-report command: `npx hyperframes feedback --search-miss "<query>" --wanted "<the move you needed>" --tier <tier>`. Wiring guidance is equally concrete (specific data attributes: data-composition-src, data-composition-id, data-start, data-duration, data-variable-values). This matches anchor 5 ('fully executable; copy-paste ready code or commands; specific examples cover the common cases'). | 5 / 5 |
Workflow Clarity | The find -> install -> wire -> edit sequence is legible through section order ("Find and install", "Wiring") with a feedback loop for search misses ("Nothing fits: report it before you hand-build") and built-in output feedback ("add ... prints the written files and the mount snippet"). However, the sequence is implied by headings rather than explicitly ordered, and there are no verification checkpoints after batch installs (`add <tag>` "installs every block with that tag") beyond the printed file list. This fits anchor 4 ('clear sequence with most checkpoints present; minor validation gaps') — above 3 because checkpoints are partially present (gap reporting, printed output, id-collision rule), below 5 because there is no explicit validate-after-install step for batch operations. | 4 / 5 |
Progressive Disclosure | SKILL.md is a genuine overview: the core loop and wiring essentials live inline, while details are pushed one level deep into clearly signaled, topically labeled references that all exist in the bundle — "references/search.md", "references/wiring-blocks.md", "references/wiring-components.md", "references/install-locations.md", and a Contributing section pointing at "references/contributing.md", "references/templates.md", "references/component-quality-bar.md" (each verified present in ./references/). References are one level deep, never nested, and each pointer states its topic. This matches anchor 5 ('clear overview with well-signaled one-level-deep references; content appropriately split; easy navigation'). | 5 / 5 |
Total | 19 / 20 Passed |