Content
72%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.
The body is an exemplary lean, well-structured workflow for a simple scaffolding skill, with concrete paths and a validation step. Its weaknesses are the unspecified core generation step delegated to @builder and the absence of an error-recovery loop after validation.
Suggestions
Add a minimal concrete example (or a referenced template) for the core artifacts of step 5 — e.g., a 3-line service/adapter snippet — so the generation step is executable rather than fully delegated to @builder.
Add a feedback loop after validation: "If typecheck or build fails, fix the errors and re-run before proceeding" rather than ending at step 7.
Clarify or define "contracts" in step 5 (where it lives, what it contains), since it is the one term not covered by the referenced ARCHITECTURE.md sections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and dense: a file reference with specific sections ("docs/ARCHITECTURE.md sections 2, 3 and 4"), a compact directory tree, terse steps, and a validation command. There is no explanation of concepts Claude already knows and nothing that could be trimmed — matching 'every token earns its place'. | 5 / 5 |
Actionability | There is concrete guidance (specific paths, directory tree, `nuxi typecheck && nuxi build`, user questions), but the core creation work — "Types + contracts + adapter + service (using $fetch)" — is delegated to `@builder` with no example or specification of what these artifacts look like, and the index.ts snippet is only comments. These are missing key details rather than the minor gaps of level 4. | 3 / 5 |
Workflow Clarity | A clear numbered sequence exists and ends with a validation checkpoint (`nuxi typecheck && nuxi build`), but there is no error-recovery loop (what to do when validation fails) and validation is deferred to the very end of a batch multi-file generation. Per the batch-operation feedback-loop cap, it cannot exceed 3. | 3 / 5 |
Progressive Disclosure | The skill is under 50 lines, single-purpose, and well-organized with a clean step structure; it appropriately defers architectural detail to the external docs/ARCHITECTURE.md rather than inlining it. Per the simple-skill guideline, this earns a 5 without bundle files. | 5 / 5 |
Total | 16 / 20 Passed |