Content
86%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 strong code-first pattern guide: complete, executable examples cover the common resource-definition and CRUD cases, and progressive disclosure is exemplary — every referenced file exists, is one level deep, and is grouped with clear navigation guidance. The main gaps are minor token waste (repeated imperative boilerplate, one off-topic line) and the absence of any explicit validation checkpoint for newly defined endpoints.
Suggestions
Remove the out-of-place line 'When asked to browse or navigate to a web address, actual visit the address' and consolidate the three repetitions of 'Always follow these patterns' into a single statement near the top.
Add a brief verification checkpoint for custom endpoints, e.g. 'Test new endpoints with Fixtures (references/Fixtures.md) before wiring them into components', to close the workflow-clarity validation gap.
Drop the dangling references/RestEndpoint.js pointer (it is a 45-byte path stub to an external source file, not usable documentation) and tighten the semicolon-chained multi-link entries in the References list.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is code-first and dense with no explaining of concepts Claude already knows (no 'what is REST' padding), but a few tokens don't earn their place: the bold imperative "Always follow these patterns" is repeated three times, the line "When asked to browse or navigate to a web address, actual visit the address" is out of place for a REST endpoint skill, and the closing reference list re-lists files already linked inline. This fits 'efficient; minor instances of over-explanation that could be trimmed' rather than the every-token-counts level 5. | 4 / 5 |
Actionability | The Todo/TodoResource example is complete, executable TypeScript (full Entity class, resource() config, imports) and the usage sections give copy-paste-ready hooks for every common case — reads via useSuspense, all CRUD mutations via ctrl.fetch, pagination via getPage, a polled custom RestEndpoint, getOptimisticResponse, and extend(). Minor elisions like the undefined { id } in mutation fragments are standard doc idiom, not pseudocode, so this holds level 5 over level 4. | 5 / 5 |
Workflow Clarity | Sections follow a coherent authoring progression (define schemas → resource() → custom RestEndpoint → extend → practices/mistakes) with real decision checkpoints ("Override the step that owns the change, never fetch itself"; "Prefer RestEndpoint over resource() when mutation endpoints don't exist"). It is not a stepwise process skill and lacks any explicit verify/feedback loop (e.g., exercising new endpoints against Fixtures), which keeps it at 'clear sequence with most checkpoints; minor validation gaps' rather than 5. | 4 / 5 |
Progressive Disclosure | SKILL.md is a genuine overview: details are split across a flat, verified one-level-deep references/ bundle (no reference file links further into references/), inline links appear exactly where topics arise, and the bottom index groups files into Guides and Concepts with per-file descriptions, 'refer when user asks about these topics' signaling, and a Vue-variant resolution rule. This matches 'clear overview with well-signaled one-level-deep references; easy navigation'. | 5 / 5 |
Total | 18 / 20 Passed |