Content
77%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.
An exemplary router skill: every trigger maps to an explicit file path, subagent, or command, with a clear decision tree and one-level-deep references throughout. Its main weakness is redundancy — the routing map is repeated in four formats (table, routing sections, two decision trees, example invocations), inflating token cost — plus a few vague or dangling lines that slightly dilute actionability.
Suggestions
Cut the Example Invocations section (or trim it to a handful of representative cases) — it duplicates the Quick Reference table and Routing Logic triggers almost verbatim, removing ~90 lines with no loss of routing information.
Keep only one decision-tree representation (the simplified numbered list) and drop the dot graph, which encodes the same branches in a harder-to-scan format.
Replace the vague "Delegate to the appropriate Cursor subagent when this router calls for a specialized auditor" with the concrete subagent list and their trigger conditions, and fix the MCP-boundary note that cites xclog/xcsym/xcprof "examples below" that do not appear in this file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body contains no filler or explanations of known concepts, but the same routing map is presented four times: the 37-row Quick Reference table, the Routing Logic trigger lists, two decision trees (dot graph plus simplified list), and ~40 Example Invocations. Roughly 40% of the body is duplicative, fitting anchor 3 ("mostly efficient but could be tightened") rather than 2 (no unnecessary concept explanation) or 4 (redundancy is substantial, not minor). | 3 / 5 |
Actionability | Mostly executable guidance: exact file paths ("See `skills/app-store-diag.md`"), concrete subagent delegations, and runnable commands ("`/axiom-triage` sentry"). However, "Delegate to the appropriate Cursor subagent when this router calls for a specialized auditor" is vague, and the MCP-boundary note references "xclog, xcsym, and xcprof examples below" that never appear in this file — minor gaps that place it at anchor 4 rather than 5. | 4 / 5 |
Workflow Clarity | As a single-task routing skill, the action is unambiguous: every decision-tree branch has labeled yes/no edges with a destination, the simplified 12-step list mirrors it, and the beta-vs-production triage split is explicitly disambiguated ("Organizer or single file" vs "multiple grouped issues"). No destructive or batch operations, so no validation cap applies; the simple-skill exception supports a 5. | 5 / 5 |
Progressive Disclosure | The SKILL.md is a pure index: all detail lives in one-level-deep, clearly signaled references ("See `skills/app-store-ref.md` (Part 11)"), each with a "Why" rationale and discovery aids (Quick Reference table). No bundle files exist to verify, and cross-suite pointers ("See axiom-security (skills/code-signing-diag.md)") remain explicit one-hop references, matching anchor 5 (clear overview, well-signaled references, easy navigation). | 5 / 5 |
Total | 17 / 20 Passed |