Content
93%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 dense, actionable process skill: every section gives concrete moves, exact file and field names, and explicit decision gates, with depth correctly deferred to family skills. The single gap is the absence of error-recovery guidance when a quality gate fails.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean throughout: the stage table compresses six stages of moves into six rows, and no line explains concepts Claude already knows — everything is operational ("Set the card's next_decision (date — decision — owner) and stop there"). The only near-duplicate, the confidence-not-cards motto, appears once as outcome framing and once as a rule, each earning its place. | 5 / 5 |
Actionability | Fully concrete for an instruction-only skill: exact paths ("reviews/YYYY-MM-DD-<slug>-decision.md"), exact card fields ("stage", "stage_entered", "next_decision"), a fully specified brief structure, named compose tools ("intent-framing-chooser", "sniff-test"), and named options ("double down / de-risk further / kill-pause"). | 5 / 5 |
Workflow Clarity | The decision-boundary sequence is numbered with explicit gates ("Only when a dated human decision exists in the card's Decision log") and verification ("confirm the flow-log row appeared"), plus a Quality Gates checklist. It falls just short of the 5 anchor because no error-recovery path exists when a gate fails — e.g., what to do if the flow-log row does not appear. | 4 / 5 |
Progressive Disclosure | Acts as a clear overview with one-level-deep, well-signaled references to depth elsewhere ("skills/sniff-test/references/ (this repo)", "Umbrella: portfolio-pdlc (leverage table, loop mechanics)") while keeping the compact stage expectations inline where they belong. No bundle files exist and no content is bloated enough to require extraction. | 5 / 5 |
Total | 19 / 20 Passed |