Content
81%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 highly actionable, well-sequenced workflow skill with executable code at every step and strong validation/feedback loops. Its weaknesses are redundancy (rules repeated across Steps and Best Practices, componentization content duplicated with its reference file) and heavy reliance on sibling-skill bundles for references.
Suggestions
State the screenshot-cadence rule and the font-assertion rule once each (e.g., in Step 5) and have Steps 1/6 and Best Practices reference them by name instead of restating them.
Replace the inlined componentization rationale in Step 4 with a two-line summary pointing to references/componentization.md, which already carries the same content.
Move the long 2a-iii inventory script and the image-transfer scripts into a local reference file (e.g., references/inventory-and-transfer-scripts.md) and keep only the call signatures and key gotchas inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly high-value, non-obvious API gotchas, but meaningful repetition: the screenshot cadence rule is stated three times (Steps 5, 6, Best Practices), the font assertion three times (Steps 1, 5, Best Practices), the parallel image-transfer workflow twice, and the componentization rationale duplicates text that also lives in references/componentization.md. Fits 'mostly efficient but could be tightened' rather than 'minor instances of over-explanation'. | 3 / 5 |
Actionability | Fully executable JS throughout — inventory script, wrapper sizing table, Promise.all import batching, setProperties text overrides with nested-instance example, createNodeFromSvg icon import with explicit viewBox/width requirements, image-hash transfer scripts, and concrete search_design_system call shapes covering common cases (variants, defaults, fonts, images). Copy-paste ready with clearly marked placeholder keys. | 5 / 5 |
Workflow Clarity | Six ordered steps with an explicit 'Do not skip steps' directive, a required-invariants block, Step 5 validation (structural read-back of descendant counts and image nodes, one composition screenshot with a targeted-fix loop, font-family assertion treated as failure on mismatch), a hard deliverable gate, and error recovery branching on safeToRetryWithoutCanvasRead. Clear sequence with explicit validation and feedback loops. | 5 / 5 |
Progressive Disclosure | Good structure: both local reference files (references/componentization.md, references/discover-product-font.md) exist, are clearly signaled with anchors, and are one level deep; a Reference Docs section indexes five figma-use references with one-line descriptions. Gaps: componentization content is duplicated between body and reference instead of delegated, several long scripts are inlined, and all but two references point into sibling skill bundles outside this skill's directory. | 4 / 5 |
Total | 17 / 20 Passed |