Conciseness

The routing table, Device Hub note, and Resources section are lean, but the ~95-line inlined session block carries substantial rhetorical padding — 'This is not negotiable. This is not optional. You cannot rationalize your way out of this', an 11-row 'Red Flags' table policing the model's own thoughts, and 'check first' repeated in nearly every row — that restates one rule many times. Mostly efficient with clear tightening opportunities; not 2 because most of that block still encodes usable routing/priority rules.

Actionability

Concrete, executable guidance throughout: 'command -v <tool>' availability check with the '<plugin-root>/tools/<tool>' fallback, a question→file routing table with named sections and a concrete command ('xcui resize sweep'), and explicit MCP tool families ('axiom_xclog_*', 'axiom_xcsym_*', 'axiom_xcprof_*'). Not 5: most tool usage is deferred to refs with no inline example invocations, and the 'Read axiom-build' target lacks a resolvable path.

Workflow Clarity

The router workflow is unambiguous — detect topic, read the named ref, follow it — with an explicit fallback chain for agent launch (MCP axiom_get_agent → file search → do the work yourself) and clear multi-domain routing examples. Not 5 because reference formats are inconsistent ('skills/xclog-ref.md' links vs. the awkward 'axiom-build (skills/build-performance.md)' pseudo-target appearing twice), which leaves minor gaps in the follow-the-instruction path.

Progressive Disclosure

The routing table and Resources section are a textbook one-level-deep router with well-signaled references, but the ~95-line behavioral session block is inlined in SKILL.md where it clearly belongs in a separate file, and the referenced skills/*.md files are not verifiable within this bundle (no references/ or skills/ directory present). Good structure overall with minor organization gaps; not 5 due to the inlined block, not 3 because navigation is easy and references are clearly signaled.

Specificity

The description lists concrete actions per tool — 'axbuild build/test capture', 'xclog console', 'xcsym .ips/MetricKit/.crash symbolication', 'xcui simulator UI/accessibility', 'xcprof xctrace/CPU profiles' — giving comprehensive, non-generic coverage of the suite's five capabilities. Not score 4: there are no gaps in per-tool action coverage; not below 5 because nothing here is vague or padded.

Completeness

The 'when' is explicit ('Use when asking about Axiom skills or axbuild...'), and the 'what' is present as compressed tool-action nouns. Not 5 because the 'what' reads as a pure trigger list rather than an explicit capability statement (no leading 'Routes/captures/symbolicates...' verb phrase); not 3 because both elements are clearly recoverable from the text itself.

Trigger Term Quality

Strong natural terms including file extensions ('.ips', '.crash', 'MetricKit', 'simulator UI', 'CPU profiles', 'symbolication', 'console'). Not 5 because several natural phrases the body itself routes here — 'screenshot a simulator', 'devicectl/simctl', 'device control', 'inspect Xcode project targets' — are missing from the description, so users asking those would not trigger it.

Distinctiveness Conflict Risk

Tool names (axbuild, xclog, xcsym, xcui, xcprof) and Axiom-specific terms form a clear niche with distinct triggers; virtually no other skill would fire on these phrases. Minimal conflict risk.