Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill reads more like a product feature description or marketing document than an actionable skill for Claude. It describes what the system should do at a high level but provides no executable code, no concrete implementation details, and no real commands. The slash-command examples reference a tool that doesn't exist without explaining how to build or invoke it, making the entire skill aspirational rather than functional.
Suggestions
Replace the conceptual descriptions with actual executable code or concrete tool invocations (e.g., show how to scan CSS files with grep/AST parsing, how to extract color values, how to generate the token JSON).
Add validation checkpoints to each mode's workflow — e.g., after scanning, verify extracted tokens are valid; after generating preview HTML, verify it renders correctly.
Either define the slash-commands as actual MCP tool calls or custom commands with implementation details, or remove them and show the actual steps Claude should take using available tools (file reading, browser MCP, etc.).
Add a concrete input/output example showing what a real design-tokens.json or audit report looks like, so Claude knows the expected format.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The content is reasonably structured but includes some unnecessary elaboration. The 10-dimension audit list and AI slop detection list are useful but could be more compact. The 'When to Use' section is somewhat redundant given the skill description. | 2 / 3 |
Actionability | The skill describes what it does conceptually but provides no executable code, no actual implementation steps, and no concrete commands that would work. The slash-command examples at the bottom appear to reference a non-existent CLI tool with no setup instructions. There's no actual code Claude could execute to perform any of these tasks. | 1 / 3 |
Workflow Clarity | The three modes have numbered steps that provide a sequence, but there are no validation checkpoints, no error handling, and no feedback loops. For a skill that involves scanning codebases and generating output files, there should be verification steps (e.g., validate generated tokens, check preview renders correctly). | 2 / 3 |
Progressive Disclosure | The content is organized into clear modes which is good, but everything is inline in one file. The outputs mention DESIGN.md and other files but there are no references to supplementary documentation. The content is borderline monolithic for its length, though the section headers help. | 2 / 3 |
Total | 7 / 12 Passed |