Content
77%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.
An exemplary thin-index skill: ordered decision flow, gated phase workflow, phase-mapped required reading, and a completion checklist, with concrete API names and thresholds throughout. The main weakness is redundancy — Core Principles and the Definition of Done restate the Core Bet and Workflow gates — and there is no inline code example. Note: the referenced rules/*.md and ../animations/ files were not present in the provided bundle (only SKILL.md), so the load-on-demand structure could not be verified against actual files.
Suggestions
State the transform/opacity + UI-thread-shared-value rule once (Core Bet or Core Principles) and cross-reference it from the Definition of Done instead of repeating it in all three sections.
Drop or compress Core Principles 4–6, which restate the Workflow phase gates (scheduleOnRN, springs for physical motion, Reduce Motion substitution) nearly verbatim — keep the gates as the single source of truth.
Add one minimal inline snippet (e.g., a shared value driven by withSpring inside a gesture callback) so the CSS-API-vs-worklet split is executable without loading a rule file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and skill-specific with no generic explanations, but the same rules are stated three to four times: Core Principles 1–2 restate the Core Bet, Principles 4–6 restate the Workflow phase gates (scheduleOnRN, springs, Reduce Motion), and the Definition of Done repeats both again. Anchor 3 ("mostly efficient but... could be tightened") fits better than 4, where over-explanation would be only minor. | 3 / 5 |
Actionability | Highly concrete directives throughout: exact API names ("animationName / transitionProperty", "withSpring / withDecay", "useNativeDriver: true", "scheduleOnRN (runOnJS on v3)"), numeric thresholds ("Duration 100–500 ms", "60/120 fps"), and tool guidance (Perf Monitor). Falls short of anchor 5's copy-paste-ready examples — there is no inline code snippet — but as a declared thin index it delegates examples to the rule files, so the gap is minor. | 4 / 5 |
Workflow Clarity | The Workflow table sequences phases A and 0–6, each with an explicit Gate (e.g., "Animated property is transform / opacity... a shared value on the UI thread"; "GestureHandlerRootView present"), the decision flow is ordered with a tie-break rule ("First match wins... pick the lower-numbered one"), the Definition of Done provides a checklist, and the measurement phase includes a diagnostic loop ("a UI-thread drop is the animation, a JS-thread-only drop is unrelated React work"). | 5 / 5 |
Progressive Disclosure | It declares itself a thin index ("Detailed rules live in rules/*.md and load on demand"), every reference is one level deep and clearly labeled, and the "Required Reading by Phase" table maps each phase to its file — a near-verbatim match for anchor 5's clear overview with well-signaled one-level-deep references. No bundle files were provided in the review directory, so scoring reflects the reference structure itself. | 5 / 5 |
Total | 17 / 20 Passed |