Content
67%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-structured router skill: no filler explanations, concrete pointers, and a clear decision tree for every networking scenario. Its weaknesses are significant redundancy (four parallel routing sections plus a per-file content re-summary) and references to `skills/*.md` files that are not present in the skill's bundle.
Suggestions
Collapse the Quick Reference table, dot-graph, numbered Decision Tree, and Example Invocations into a single routing structure — the dot graph in particular renders nowhere and duplicates the numbered tree.
Trim or remove the Critical Patterns section, which re-summarizes the contents of files the reader is about to open; the table's pointers already establish scope.
Verify the `skills/*.md` reference paths resolve from the skill's own bundle (or move them under references/), since none of the referenced files exist alongside this SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body never explains concepts Claude already knows, but the same routing is repeated four ways — an 18-row Quick Reference table, a `dot` graph, a 10-item numbered decision tree, and Example Invocations — plus a Critical Patterns section that re-summarizes each referenced file; it could be tightened considerably. Version-sensitive details ("iOS 26+", "iOS 12-18", "WWDC 2025") add date-bound tokens, though legacy content is at least sectioned off. | 3 / 5 |
Actionability | Pointers are concrete and specific — exact file paths ("skills/networking-diag.md"), named agents ("networking-auditor"), and exact commands ("/skill axiom-security", "/axiom-audit networking") — but the body itself contains no executable networking guidance, and the referenced `skills/*.md` files are absent from this bundle, so the pointers cannot be verified where the skill lives. | 4 / 5 |
Workflow Clarity | The routing sequence is unambiguous and covers edge cases (security delegation, watchOS limits, deprecated-API audits), and the pressure-resistance steps ("Diagnose first → Recommend correctly → Respect but don't enable") are clearly ordered; the gaps are minor — duplicated routing structures could produce ambiguity and there is no guidance on what to do after reading a reference. No validation cap applies since the skill involves no destructive or batch operations. | 4 / 5 |
Progressive Disclosure | The router form is textbook — SKILL.md is a lean overview with clearly signaled, one-level-deep references organized in a symptom-to-file table — but the bundle contains no references/, scripts/, or skills/ directory, so every referenced path (skills/networking-discipline.md etc.) is unverifiable within the skill's own bundle, leaving minor-to-real organization gaps versus the anchor-5 fully self-contained structure. | 4 / 5 |
Total | 15 / 20 Passed |