Content
71%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: symptom-based sections, a prioritized decision tree, and one-level-deep references to per-topic files make it highly navigable and actionable. Its main cost is token redundancy — the Decision Tree and Example Invocations sections substantially duplicate the Routing Logic sections and could be consolidated into a single routing surface.
Suggestions
Consolidate the Decision Tree and Example Invocations sections into the Routing Logic sections (or vice versa): the same ~30 routes are stated three times, and merging them into one routing surface would cut roughly a third of the body without losing any route.
Make the 'Cursor Subagent Routing' section concrete: list the actual subagent names (memory-auditor, energy-auditor, swift-performance-analyzer, triage-analyzer, performance-profiler) and when each is preferred over the file-based route, or remove the section since it duplicates the routing entries below.
Add misdiagnosis feedback loops to the Decision Tree (e.g., 'if hang-diagnostics rules out a busy main thread, cross-route to performance-profiling') so a wrong first route has a stated recovery path rather than dead-ending.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Individual sections are tight and free of concept over-explanation, but the same routing information appears three times — the Routing Logic sections, the 27-item Decision Tree, and ~130 lines of Example Invocations largely restate one another, which is more than minor trimmable redundancy. It is not a 2 because no section is padded with explanations of concepts Claude already knows. | 3 / 5 |
Actionability | Nearly every route names a concrete executable target: exact file paths ("skills/memory-debugging.md"), named agents ("memory-auditor"), commands ("/axiom-profile"), and an explicit MCP tool mapping with fallback behavior at the top. The gap is the "Cursor Subagent Routing" section, which is a vague stub ("Delegate to the appropriate Cursor subagent") that specifies neither which subagents nor when. | 4 / 5 |
Workflow Clarity | The Decision Tree provides a prioritized, sequenced diagnostic workflow with explicit ordering ("memory-debugging FIRST… then performance-profiling if memory is fixed but stutter remains") and symptom-based branching. No validation checkpoints exist, but as a pure routing skill it involves no destructive or batch operations, so the validation cap does not apply; it falls short of 5 because there are no feedback loops for misdiagnosis (e.g., what to do when the routed skill's fix doesn't resolve the symptom). | 4 / 5 |
Progressive Disclosure | The body is a pure index: every topic's detail lives in a separate per-topic file, all references are one level deep, clearly signaled ("See skills/X.md"), and organized under symptom-based headings for easy navigation — closely matching the well-signaled one-level-deep reference anchor. No bundle files exist in this skill, and no nested or buried references were found. | 5 / 5 |
Total | 16 / 20 Passed |