Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill provides useful output templates for three design system tasks but is significantly bloated with definitional content Claude already knows (what tokens, components, and patterns are). It lacks concrete, step-by-step workflows for how to actually perform each task — particularly the audit, which needs specific instructions on what to scan and how. The output templates are valuable but would be better served as referenced files rather than inline content.
Suggestions
Remove the entire 'Components of a Design System' section — Claude already knows what design tokens, components, and patterns are. This saves ~20 lines of unnecessary context.
Add concrete step-by-step workflows for each subcommand, especially 'audit': specify what files to scan (CSS, component files), what patterns to grep for (hardcoded hex values, arbitrary spacing), and how to compile findings.
Move the three large output templates to separate referenced files (e.g., AUDIT_TEMPLATE.md, DOCUMENT_TEMPLATE.md, EXTEND_TEMPLATE.md) and keep only a brief summary of each output format in the main skill.
Add validation steps to the audit workflow — e.g., verify findings against actual token definitions before reporting, confirm component file paths exist before documenting.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is extremely verbose at ~150+ lines, with large sections explaining what design tokens, components, and patterns are — concepts Claude already knows well. The 'Components of a Design System' section is pure definitional content that adds no actionable value. The three full output templates are lengthy and could be more concise. | 1 / 3 |
Actionability | The output templates provide concrete markdown structures that serve as useful scaffolding, and the usage commands are clear. However, there are no executable code examples, no concrete steps for how to actually perform an audit (e.g., what files to scan, what regex patterns to use for hardcoded values), and the guidance remains at the template/description level rather than providing specific executable instructions. | 2 / 3 |
Workflow Clarity | The three subcommands (audit, document, extend) are clearly separated with distinct output templates, but there's no step-by-step workflow for any of them. For example, the audit command doesn't specify: scan these file types → look for these patterns → compile results → validate findings. There are no validation checkpoints or feedback loops for what could be a complex multi-step process. | 2 / 3 |
Progressive Disclosure | The skill references CONNECTORS.md appropriately, and the content is organized into logical sections. However, the three massive output templates bloat the main file significantly — these could be split into separate reference files. The 'Components of a Design System' definitional section is inline content that doesn't belong in the main skill file at all. | 2 / 3 |
Total | 7 / 12 Passed |