Content
71%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.
Highly actionable with executable commands and complete payloads, supported by a clear loop with feedback checkpoints. Weaker on token efficiency (repetition) and progressive disclosure (monolithic inlining of material that could live in separate files).
Suggestions
De-duplicate the three-exit guidance: state it once in the loop section and have step 5 reference it instead of restating it.
Split the 21-recipe list and the motion-token tables into reference files under references/ and link to them, keeping SKILL.md an overview with the core loop.
Add an explicit verification checkpoint in the Apply job path (e.g. confirm the edited file still parses / the selector still matches) before posting applied:true.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense and instruction-rich without teaching known concepts, but the three exits, the start-latch logic, and the '/refine live' resume guidance are each restated multiple times and could be tightened. | 3 / 5 |
Actionability | Every endpoint has a copy-paste curl command with a full JSON body, plus concrete token tables and a complete recipe list covering the common cases. | 5 / 5 |
Workflow Clarity | Clear numbered loop with explicit feedback checkpoints (stop signal, idle backoff with reset, result vs error); the destructive apply path lacks an explicit edit-validation step, leaving a minor gap. | 4 / 5 |
Progressive Disclosure | Well-sectioned but ~450 lines with bulk reference material (21-recipe list, motion-token tables, scan/apply specs) inlined; no bundle files exist and the named reference paths point outside this skill's bundle. | 3 / 5 |
Total | 15 / 20 Passed |