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 dense, highly actionable body with concrete commands, named tools, and specific guardrails, supported by a clear scene-based workflow with verification and recovery steps. Its main weaknesses are meta-process scaffolding that inflates token cost without executable value, and reliance on external resource files that are not present in the provided bundle.
Suggestions
Cut the "Actions | SSL primitive" table and the PREPARE/ACQUIRE/ACT/VERIFY/FINALIZE scene ceremony; keep only the entry/exit and verification steps that change behavior.
Remove the "Intent signature" and "When to use" sections — they duplicate the frontmatter description and the routing decision.
Defer the shadcn engine and library-rationale detail (already owned by `resources/tech-stack.md`) out of SKILL.md, keeping only the decision table itself.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body mostly assumes Claude's competence (no React/PDF-style concept explanations, dense imperative tables), but carries unnecessary meta-scaffolding: the "Actions | SSL primitive" table mapping to generic `READ`/`SELECT`/`NOTIFY` primitives, the PREPARE/ACQUIRE/ACT/VERIFY/FINALIZE "Scenes" ceremony, and "Intent signature"/"When to use" sections that restate the frontmatter description. This fits anchor 3 — mostly efficient but could be tightened — rather than 4, because multiple sections add tokens without adding executable guidance. | 3 / 5 |
Actionability | Largely executable: concrete `rg` commands in the canonical workflow path, named shadcn registry tools, a Libraries table with specific versions and decision rules, and guardrails with copy-paste-ready specifics (`prefetch={false}`, `order-summary-card.tsx`, `TestScheduler` marble tests). Not a 5 because verification commands are delegated generically ("lint, typecheck, tests... when applicable") rather than given, and some guidance ("Map code variables to Figma layer names") is direction without a step. | 4 / 5 |
Workflow Clarity | A clear Entry → PREPARE/ACQUIRE/ACT/VERIFY/FINALIZE → Exit sequence with an explicit verification scene, a "run the checklist before handoff" checkpoint, and a Failure-and-recovery feedback loop ("If verification fails, fix before handoff or report the blocker"). Not a 5 because the sequence is phrased at the scene level and the actual step-by-step is deferred to `resources/execution-protocol.md`, which is not present in the bundle to confirm. | 4 / 5 |
Progressive Disclosure | References are one level deep, clearly signaled, and well organized — the Stack Reference table states what each `resources/*.md` file "Owns" and the References section annotates every pointer's purpose. Not a 5 because the bundle contains no `references/`, `scripts/`, or `assets/` directories, so the referenced `resources/*.md`, `../../rules/frontend.md`, and `../_shared/core/*` files cannot be verified to exist, and some inline content (the shadcn engine details and Libraries rationale) overlaps with `resources/tech-stack.md` rather than deferring to it. | 4 / 5 |
Total | 15 / 20 Passed |