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, highly actionable read-only status skill with a clear phased workflow and concrete command routing for every recommendation. The main drag on conciseness is the long illustrative output block and a repeated verbose judgment-caveat pattern that could be DRY'd.
Suggestions
Factor the repeated '默认参考 / 但 Claude 可以更早 / 也可以更晚 / 提示时显式标注' block (triggers #7 and #19) into a single shared note referenced once, instead of repeating it verbatim.
Shorten the Phase 4 illustrative dashboard to the essential format skeleton plus one short example, rather than a full ~35-line mock output.
Replace the Phase 1/2 pseudocode helpers (read_json, parse_candidates_md_entries, wc -l) with concrete executable snippets or explicit file-reading commands so the internal logic is as actionable as the output recommendations.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is operational and avoids explaining concepts Claude already knows, but the ~35-line illustrative dashboard output and the verbatim '默认参考/但 Claude 可以更早/也可以更晚' caveat block repeated in triggers #7 and #19 could be tightened or factored out. Not 4 because the repeated verbose caveat block is noticeable padding; not 2 because the content is genuinely operational rather than padded with generic explanation. | 3 / 5 |
Actionability | Every trigger condition routes to a concrete copy-paste command (e.g. '/cheat-retro predictions/...', '/cheat-bump --propose', '/cheat-seed', '/cheat-trends'), and Key Rule 4 enforces this. Not 5 because Phase 1/2 use pseudocode helpers (read_json, glob, parse_candidates_md_entries, wc -l) that are not executable as written; not 3 because the user-facing deliverable commands are fully concrete. | 4 / 5 |
Workflow Clarity | Four phases are clearly sequenced (read state → derive metrics → detect triggers → output), triggers are priority-ordered 1-19, and there is a missing-state checkpoint plus the '不假装数据可用 → 标未知' validation rule. Not 5 because there is no error-recovery feedback loop (acceptable for a read-only skill but keeps it off the top anchor); not 3 because checkpoints are explicitly present. | 4 / 5 |
Progressive Disclosure | Well-organized into Overview/Constants/Inputs/Workflow/Key Rules/Refusals/Integration, with two clearly-signaled one-level references ([cadence-protocol.md], [state-management.md]) inline as markdown links; the 19 trigger rules are correctly inlined since they are needed at runtime. Not 5 because no bundle files exist and the bulk of content lives inline with only two external references; not 3 because structure and signaling are clear. | 4 / 5 |
Total | 15 / 20 Passed |