Content
63%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-constructed router: genuinely useful, version-sensitive routing knowledge with concrete per-scenario file destinations and a clear triage-first sequence. Its weaknesses are substantial cross-section redundancy (the same routes appear in up to five sections) and a progressive-disclosure structure that leans entirely on sibling skill files not present in this bundle.
Suggestions
Consolidate the routing information that currently appears in five places (AI Approach Triage, Routing Logic, Decision Tree, Critical Patterns, Example Invocations) into one canonical routing table plus a short worked-examples section; drop the redundant 'Critical Patterns' section that merely re-lists foundation-models and foundation-models-diag bullets.
Ship the most-load-bearing referenced files (e.g., a foundation-models quick-reference) inside this skill's references/ directory, or restate the handful of critical facts (Approach Triage rungs 1-4, the adapter OS-27 obsolescence, the SpeechAnalyzer audio-session trap) inline so the skill is self-sufficient when sibling skills are unavailable.
Make cross-skill invocations actionable by specifying the mechanism ('invoke /skill axiom-concurrency', 'read axiom-security/skills/agentic-security.md') instead of bare 'also invoke axiom-security' instructions with no path or command.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-obvious, valuable specifics (OS-version constraints, entitlements, bug IDs like FB18924722) and never re-teaches concepts Claude knows, but the same routing decisions are restated five times across 'AI Approach Triage', 'Routing Logic', 'Decision Tree', 'Critical Patterns', and 'Example Invocations' — the overlap could be consolidated into one table plus one example section. | 3 / 5 |
Actionability | Routing is concrete and executable for its purpose: exact file paths per scenario, a numbered 14-step decision tree, and ~20 worked user-utterance → file mappings ('My AI generation is being blocked' → 'skills/foundation-models-diag.md'). Minor gaps keep it below 5: instructions like 'also invoke axiom-concurrency' and 'axiom-security' give no invocation mechanism, and several routes are duplicates of the triage table. | 4 / 5 |
Workflow Clarity | The routing workflow is clearly sequenced: 'First, determine which kind of AI the developer needs', then triage tables, training-path boundaries, a numbered decision tree, and worked examples. No validation/feedback checkpoints exist, but the skill routes rather than performs destructive or batch operations, so this is a minor gap rather than a cap. | 4 / 5 |
Progressive Disclosure | References are well-signaled and one level deep (each 'skills/*.md' pointer is inline and labeled), but the bundle contains none of them — there is no references/ directory and files like 'skills/foundation-models.md' and 'skills/speech.md' are absent from this skill's bundle, so the disclosure structure is entirely external to the skill and cannot be navigated from what ships here. | 3 / 5 |
Total | 14 / 20 Passed |