Content
92%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 procedure-style skill: a tightly sequenced five-step workflow with explicit stop conditions, validation, and error-recovery guidance, concrete API-level directives throughout, and clean delegation of detail to an existing, purpose-conditioned bundle. The only real weakness is redundancy between the Procedures and Error Handling sections, where three rules are stated multiple times.
Suggestions
Deduplicate the 'downloading before create()' guidance, which appears in Step 3 item 7, Step 4 item 6, and Error Handling bullet 2 — state the rule once and cross-reference it from the other sites.
The Error Handling section largely restates procedure content (iframe allow, Web Worker limitation); keep only the genuinely new recovery guidance there (NotSupportedError alignment, polyfill fallback) and trim the repeats to save tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense, procedural, and free of concept explanation Claude already knows — every step is a directive, not a tutorial. However, several rules are stated two or three times: the "downloading before create()" guidance appears in Step 3 item 7, Step 4 item 6, and Error Handling bullet 2; iframe `allow="language-model"` appears in Step 3 item 5 and Error Handling bullet 5; the Web Worker limitation appears in Step 2 item 6 and Error Handling bullet 4. This matches the anchor for efficient content with minor instances that could be trimmed, not level 5 where every token earns its place. | 4 / 5 |
Actionability | Guidance is fully concrete and directive throughout: an executable command (`node scripts/find-frontend-targets.mjs .`), exact file reads with conditions (`references/prompt-api-reference.md`, `assets/language-model-service.template.ts`), exact API methods and options (`prompt()`, `promptStreaming()`, `responseConstraint`, `measureContextUsage()`), and explicit do/don't rules ("Do not pass `tools` to `availability()`", "Do not depend on `params()`, `topK`, or `temperature`"). Code is delegated to a real 381-line template asset rather than inlined, which is appropriate for an instruction-driven skill; the guidance itself is specific and covers the common cases. Matches the fully actionable anchor. | 5 / 5 |
Workflow Clarity | A clear five-step sequence (identify surface → confirm viability → implement wrapper → wire UX → validate) with explicit checkpoints and feedback loops: stop-and-ask conditions ("stop and ask the user which app should receive the integration", "stop and explain that this skill does not apply"), a validation step ("Run the workspace build, typecheck, or tests after editing"), and an Error Handling section with recovery redirects (NotSupportedError handling, worker redirect, iframe requirement). Matches the anchor for clear sequence with explicit validation steps and feedback loops for error recovery. | 5 / 5 |
Progressive Disclosure | The SKILL.md body is a ~50-line procedural overview that delegates all detail to one-level-deep, well-signaled bundle files, each referenced with a purpose condition: "Read `references/prompt-api-reference.md` before writing code", "Read `references/examples.md` when the feature needs a spec-valid message shape", "Read `references/compatibility.md` when the feature must support multiple browser generations", plus `polyfills.md`, `troubleshooting.md`, the discovery script, and the service template — all verified to exist. Matches the anchor for a clear overview with well-signaled one-level-deep references and easy navigation. | 5 / 5 |
Total | 19 / 20 Passed |