Conciseness

There is no conceptual padding or explanation of things Claude already knows, but the routing map is restated in four overlapping representations — Quick Reference table, dot-graph Decision Tree, numbered Decision Tree list, and Example Invocations (e.g. Siri → skills/app-intents-ref.md appears in all of them). Anchor 3 ('mostly efficient but could be tightened') fits; the redundancy is more than the 'minor trimming' of anchor 4, though the Anti-Rationalization table earns its tokens with non-obvious facts.

Actionability

Every row ends in an exact executable action — "See `skills/push-notifications-diag.md`", "Launch `iap-auditor` agent", "Read: `skills/siri-audio.md`" — and Example Invocations map common user phrasings verbatim to files. For an instruction-only router this is the equivalent of copy-paste-ready commands, matching anchor 5.

Workflow Clarity

The classify-then-route procedure is unambiguous (numbered tree with 1a/1b/1c sub-branches), and Cross-Domain Routing plus Conflict Resolution explicitly resolve the genuinely ambiguous cases ("stay here" vs "also invoke axiom-media", integration vs axiom-build). Minor gaps keep it at anchor 4: no in-router fallback when no row matches, and no stated precedence among the three overlapping routing representations.

Progressive Disclosure

The body is a pure index: ~40 one-level-deep references (`skills/<name>.md`), each row signaling exactly what it covers (e.g. "Live Activities / ActivityKit API reference → skills/live-activities-ref.md"), with no detail content inlined — a textbook match for anchor 5. Caveat: the referenced skills/*.md files are not present in this review bundle, so paths were verified structurally rather than on disk.

Specificity

The description enumerates ~20 concrete system surfaces ("Siri and Shortcuts, widgets, Live Activities, in-app purchase, localization, privacy prompts, alarms, timers, calendar, reminders, contacts, SharePlay, VoIP and CallKit..."), which is far richer domain naming than anchor 2's example, but it states no actions or capabilities whatsoever, falling short of anchor 3's '1-2 concrete actions'.

Completeness

The 'when' is explicit and maximally detailed ("Use when work crosses into Apple's system surfaces..."), but no 'what' — no statement of what the skill does — appears anywhere. This is the mirror of anchor 3 (one half clear, the other missing/absent), clearly above anchor 2's vague 'when' example but below anchors 4-5 which require both halves.

Trigger Term Quality

Highly natural user terms across ~20 domains ("alarms", "timers", "calendar", "reminders", "contacts", "weather", "push", "privacy prompts"), giving good-to-excellent keyword coverage. A few natural terms/synonyms are missing ("push notifications" spelled out, "subscriptions", "extensions"), so it fits anchor 4 rather than anchor 5's comprehensive synonym coverage.

Distinctiveness Conflict Risk

"rather than your own UI or data" is an explicit disambiguator against sibling suites, and the trigger list carves a clear niche of Apple system surfaces. Matches anchor 5: clear niche with distinct triggers and minimal conflict risk.