Content
86%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 lean, well-structured workflow skill that assumes Claude's competence and gives a crisp numbered procedure with explicit priority ordering and error-recovery feedback loops. The only gaps are the absence of a concrete adaptation example and of a final output-verification step, which keep actionability and workflow clarity just short of full marks.
Suggestions
Add one brief concrete example of adapting reference output to a target project — e.g., a before/after snippet showing a Tailwind class from get_design_context mapped to the project's token or component — to lift actionability.
Add a final verification step to the workflow, such as comparing the implemented screen against the screenshot returned by get_design_context and re-checking hint priorities, to close the validation gap.
Trim the 'Direction and Scope' section: the intro paragraph already states the read-FROM-Figma direction and the reverse-direction pointer, so the two MUST bullets are largely redundant and could be merged into one line.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~44-line body is dense imperatives with zero concept explanations: 'You MUST call get_design_context on the target node before writing any code', 'Apply the hints in this order — earlier sources override later ones'. It even delegates parameter mechanics to the tool description ('live on the get_design_context tool description itself — follow them there') instead of restating them. Every token earns its place. | 5 / 5 |
Actionability | Concrete, executable procedure throughout — tool call order, bans on substituting get_metadata/get_screenshot, a five-level hint-priority list, and specific error-recovery steps ('ask the user for a node-specific URL', 'retry against a smaller node'). Not 5: there are no concrete examples of the core adaptation step (e.g., a hint-conflict case or a snippet mapping reference output to a project's stack), leaving the trickiest step abstract. Not 3: the guidance is specific and executable, not high-level hints. | 4 / 5 |
Workflow Clarity | A numbered four-step workflow (call → treat as reference → reuse → honor hints by priority) with an Error Recovery section providing feedback loops (stop-and-read on error, retry smaller node, ask user for missing node-id). Not 5: there is no final validation of the produced code, such as checking the implementation against the returned screenshot — the checkpoints cover tool errors but not output verification. Not 3: the sequence and error checkpoints are explicit, not implicit. | 4 / 5 |
Progressive Disclosure | No bundle files exist, and none are needed: the under-50-line body is well-sectioned (Direction and Scope, Workflow, Error Recovery), appropriately pushes parameter mechanics to the get_design_context tool description, and points reverse-direction work to the figma-generate-design skill. Fits the well-organized-sections exception cleanly. | 5 / 5 |
Total | 18 / 20 Passed |