Content
45%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 a pure routing hub with a well-designed navigation scheme — symptom table, decision tree, cross-skill conflict resolution, and anti-rationalization guidance — but two problems dominate. First, the routing information is duplicated across four largely redundant sections, wasting tokens. Second, every referenced skills/*.md file is absent from the bundle, so the skill routes to nothing and provides no executable guidance itself.
Suggestions
Eliminate the duplicate routing sections: the dot decision tree restates the numbered decision tree, and Example Invocations restates the Quick Reference table. Keep the Quick Reference table plus one decision tree and cut the rest (~60 redundant lines).
Fix the bundle: add the six referenced skills/*.md files (core-location, core-location-ref, core-location-diag, mapkit, mapkit-ref, mapkit-diag) — ideally under the standard references/ directory — or inline minimal quick-start guidance so the routing targets actually resolve.
Include at least a minimal inline executable example (e.g., a basic CLLocationUpdate listener or SwiftUI Map snippet) so the skill remains actionable when the referenced files are unavailable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The same routing information is repeated four times: the 17-row Quick Reference table, the dot-format decision tree, the numbered decision tree (which restates the dot graph exactly), and the 9-entry Example Invocations section (which restates the Quick Reference), with Critical Patterns partially duplicating it a fifth time. This is noticeably verbose with several padded, redundant sections — roughly 60 of 165 lines are duplicate re-routings — though it never pads with concepts Claude already knows, so it stays above a 1. | 2 / 5 |
Actionability | Concrete per-symptom routing directives are written ('Location updates never arrive' → skills/core-location-diag.md), and the Anti-Rationalization table adds specific facts. But the guidance is incomplete in practice: none of the six referenced skills/*.md files exist in the bundle, so every 'Read X' directive dead-ends, and there is no inline executable fallback (no code, commands, or quick-start). | 3 / 5 |
Workflow Clarity | The decision sequence is clear and unambiguous: task → Core Location vs MapKit → strategy/API/diagnostics → specific file, reinforced by explicit precedence rules in Conflict Resolution ('Try location FIRST', 'Only use axiom-performance if...'). No validation checkpoints are needed since the skill routes rather than performs destructive or batch operations; it falls short of 5 only because there is no canonical single path — the redundancy slightly obscures which section is authoritative. | 4 / 5 |
Progressive Disclosure | The reference signaling design is excellent — one-level-deep, symptom-indexed, consistent paths — but the bundle contains zero files: none of the six referenced skills/*.md files (core-location, core-location-ref, core-location-diag, mapkit, mapkit-ref, mapkit-diag) exist, and there is no references/ or skills/ directory. The disclosure chain is broken: the SKILL.md is a router to a missing book, leaving no substantive content reachable. | 2 / 5 |
Total | 11 / 20 Passed |