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 diagnostic router with concrete, executable routing targets and exemplary one-level-deep progressive disclosure. Its main weakness is redundancy: Routing Logic, Decision Tree, Critical Patterns, and Example Invocations restate the same symptom-to-skill mappings, inflating token cost by roughly a third, with one internal inconsistency ("5 common leak patterns" in Routing Logic vs "6 leak patterns" elsewhere).
Suggestions
Cut or drastically trim the Example Invocations section — nearly every entry duplicates a Decision Tree or Routing Logic row (e.g., memory leak → memory-debugging appears in all three), so keeping only mappings not already covered would reclaim ~100 lines.
Merge the Critical Patterns section into the Routing Logic bullets it paraphrases, which would also eliminate the leak-pattern count inconsistency ("5 common leak patterns" at line 40 vs "6 leak patterns" at lines 43, 224, and 249).
Specify how to resolve cross-suite references like "`axiom-shipping (skills/production-triage.md)`" and "axiom-build" (root path or lookup rule), as is already done for agent files in the inline auditor block, so every route is unambiguous.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense routing information with no explanations of concepts Claude already knows, but the same mappings are stated three times: Routing Logic, the 27-item Decision Tree, and a ~130-line Example Invocations section repeat "memory leak → memory-debugging" style routes verbatim, and Critical Patterns restates Routing Logic summaries. This fits 'mostly efficient but could be tightened' rather than 2, since the padding is duplication of genuinely useful routing rather than fluff or concept tutorials. | 3 / 5 |
Actionability | Routes are concrete and executable: exact file paths ("skills/memory-debugging.md", "skills/hang-diagnostics.md"), copy-paste commands and slash commands ("/axiom:audit memory", "/axiom:profile", "/axiom:console", "xcprof compare --fail-on-regression"), and specific flags ("--timeout", "--max-lines"). Not 5 because cross-suite references like "See `axiom-shipping (skills/production-triage.md)`" never specify how to resolve that suite's root path, leaving the reader to guess; everything else is directly actionable. | 4 / 5 |
Workflow Clarity | The Decision Tree provides a clearly sequenced diagnostic flow with explicit YES/NO branches, prioritization ("memory-debugging FIRST ... then performance-profiling if memory is fixed"), and a verification loop via "Compare two traces / detect regressions" with non-zero exit semantics. Not 5 because most leaf routes end at 'see file' with no checkpoint confirming the diagnosis before acting, and there is no explicit error-recovery step if a route yields nothing. | 4 / 5 |
Progressive Disclosure | This is a textbook router: the SKILL.md body contains only routing (topic sections, a decision tree, an anti-rationalization table) and defers all detail to one-level-deep, clearly signaled references ("See skills/performance-profiling.md", "See axiom-build (skills/lldb.md)"). No nested references — target files are leaf skills, not further routers. No bundle files exist in this skill's directory, and the body appropriately pushes all substance to the referenced sibling skills. | 5 / 5 |
Total | 16 / 20 Passed |