Content
70%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured troubleshooting skill with excellent workflow clarity and progressive disclosure. Its main weaknesses are verbosity in the anti-patterns/philosophy/when-to-use sections (which explain reasoning Claude already possesses) and a lack of concrete, executable examples — the usage examples are pseudocode comments rather than actionable demonstrations. The duplicated references section also wastes tokens.
Suggestions
Remove or significantly condense the 'When to Use', 'When Not to Use', 'Anti-Patterns', and 'Philosophy' sections — these explain reasoning Claude already understands and consume significant tokens.
Replace the pseudocode usage examples with concrete, executable examples showing actual commands, search queries, or diagnostic outputs that Claude would produce.
Consolidate the duplicated 'Refs' and 'References' sections into a single section to save tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is mostly efficient with good use of shorthand and structured steps, but includes some redundancy: the 'When to Use'/'When Not to Use'/'Anti-Patterns' sections are verbose and explain reasoning Claude already understands. The Philosophy section and Anti-Patterns explanations with 'Why:' annotations add bulk. The Refs section is duplicated (listed twice with slightly different formatting). | 2 / 3 |
Actionability | The workflow steps provide structured guidance but lack concrete executable examples. The 'Usage Examples' section shows only comments/pseudocode rather than actual commands or code. Key steps like 'WebSearch: [error] [stack] [framework]' are templates rather than executable instructions. The AskUserQuestion guard is concrete and actionable, but most diagnostic steps remain abstract. | 2 / 3 |
Workflow Clarity | The workflow is clearly sequenced (0-6) with explicit validation checkpoints: search before diagnosing, reproduce before theorizing, confirm root cause before persisting, and the OODA loop has clear exit conditions. The completion checklist and mandatory persist step with preconditions demonstrate good feedback loops. The AskUserQuestion guard adds an important error-recovery mechanism. | 3 / 3 |
Progressive Disclosure | The skill provides a clear overview with well-signaled one-level-deep references to diagnose.md, search-multi-source.md, and reference.md. Step 3 explicitly defers to 'references/protocols/diagnose.md for details.' The References section clearly describes what each linked file contains. Content is appropriately split between the overview and detailed reference files. | 3 / 3 |
Total | 10 / 12 Passed |