Content
27%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill is extremely verbose and reads more like product documentation for a hypothetical tool than an actionable skill for Claude. It contains significant amounts of aspirational content (CLI flags, CI/CD integration, baseline tracking) that don't correspond to real capabilities, and the core mechanism of spawning parallel subagents lacks concrete implementation details. The referenced bundle files don't exist, and the massive inline content would benefit greatly from being split across multiple files.
Suggestions
Reduce content by 60-70%: remove the architecture benefits list, troubleshooting section, best practices, and CI/CD integration. Focus on the core workflow of spawning subagents with specific instructions.
Make the subagent spawning actionable: provide the exact tool calls or API patterns Claude should use to launch parallel subagents, rather than describing an abstract architecture.
Split detection patterns, security checks, and complexity thresholds into referenced files (e.g., references/patterns.md) and actually include them in the bundle.
Add validation checkpoints: verify subagent completion, validate that auto-fixes don't break tests, and include error recovery for failed scans.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Extremely verbose at ~350+ lines. Contains extensive explanatory content Claude doesn't need (architecture diagrams with emoji, benefit lists, detailed explanations of what each scanner does). The report template, CI/CD integration, baseline tracking, custom patterns, troubleshooting, and best practices sections add massive token overhead. Much of this is aspirational documentation rather than actionable instruction. | 1 / 3 |
Actionability | Provides structured guidance with specific patterns, thresholds, and report formats, but the core mechanism relies on fictional CLI commands (/techdebt) and tools that don't exist. The subagent instructions template is useful but the actual implementation details for spawning subagents and using tools like ast-grep are vague. Pre-commit hooks and CI/CD examples reference non-existent 'claude skill techdebt' commands. | 2 / 3 |
Workflow Clarity | The 5-step workflow is clearly sequenced and includes consolidation/deduplication steps. However, there are no validation checkpoints for the scanning process itself — no verification that subagents completed successfully, no error handling for failed scans, and the auto-fix mode lacks a validation step after applying fixes to confirm the code still works. | 2 / 3 |
Progressive Disclosure | References two bundle files (references/patterns.md, references/severity-guide.md) that don't exist in the bundle. The skill is a monolithic wall of text with everything inline — detection patterns, language support tables, integration patterns, advanced usage, troubleshooting, and best practices all crammed into one file when much of this content should be split into referenced files. | 1 / 3 |
Total | 6 / 12 Passed |