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 well-structured routing skill: modes table, explicit load triggers, verified one-level-deep references, golden rules, and a closing feedback loop. The main costs are token inefficiency from boilerplate repeated across the three style sections and abstract directives whose substance lives entirely in the reference files.
Suggestions
State the shared four-step workflow (identify repos → inspect → adapt → return) once in a single section and have each style section link to it, cutting ~12 repeated lines.
Merge the "Did this help?" section into the Quality Checklist item that already mandates the closing question to remove the duplication.
Make the abstract directives executable, e.g. replace "produce output that feels like it came from that style" with the concrete check the reference applies (name key repos, mirror naming/structure conventions, show a before/after snippet).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient (tables, terse principle bullets, no explanations of concepts Claude already knows), but the identical four-step "How To Use This Style" block ("Identify relevant repos from the map / Inspect them (clone to /tmp...) / Adapt patterns responsibly / Return to the user's codebase...") is repeated nearly verbatim in all three style sections, and the "Did this help?" section duplicates a Quality Checklist item. This matches "mostly efficient but could be tightened"; not a 4 because the triplicated boilerplate is a clear trim target, not a minor instance. | 3 / 5 |
Actionability | As an instruction/routing skill it gives concrete executable guidance: exact reference paths per mode in the Modes table ("references/shadcn.md"), bold load directives with specific trigger conditions, and a defined per-style procedure (identify repos from the map, clone to /tmp, adapt, return to the codebase). Not a 5 because some directives stay abstract ("produce output that feels like it came from that style", "adapt patterns responsibly") with the concrete specifics fully deferred to references; not a 3 because the routing and steps are specific and executable, not high-level hints. | 4 / 5 |
Workflow Clarity | The multi-step process is clearly sequenced: pick a mode from the table, load the matching reference, follow its "How To Use This Style" steps, apply the Golden Rules, then run the Quality Checklist, ending with a feedback question ("Did this solve what you were trying to do?"). The checklist provides checkpoints and the closing question a feedback loop. Not a 5 because there are no error-recovery loops (e.g. what to do when a repo can't be inspected or a pattern doesn't fit) and the sequence is spread across sections rather than stated once. | 4 / 5 |
Progressive Disclosure | Scored against the actual bundle: references/shadcn.md, references/sindresorhus.md, and references/ahmadawais.md all exist with the sections the body cites ("How To Use This Style", repo maps), references are one level deep, and they are clearly signaled via the Modes table's Load column and bold "Always load" directives. Not a 5 because the inline "Key principles from the reference (internalize these)" lists duplicate reference content that belongs behind the disclosure boundary, and the routing table's "Relevant reference(s)" / "Load target reference + current code" entries are less precise than the named files. | 4 / 5 |
Total | 15 / 20 Passed |