Content
73%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.
The content is a well-sequenced, heavily validated workflow with concrete commands and hard gating rules — its strongest qualities. Its weaknesses are token efficiency (over-decomposed micro-steps and long enumerations that could be compressed) and an inlined environment-specific preview section that would fit better in an external reference.
Suggestions
Compress the capture micro-steps ("Start at the top... Scroll down in small steps... Continue until the full page has been seen") and the 15-item DOM enumeration into a few dense directives that assume Claude's competence.
Move the "Previewing prototypes in ChatGPT Work Mode" section into local-prototype-preflight.md or a dedicated reference, keeping only a one-line pointer in SKILL.md.
Specify how to invoke ImageGen for asset replacement (or link to a reference that does), since it is currently named without any actionable detail.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body avoids concept explanations, but micro-instructions like "Start at the top of the page. Scroll down in small steps. At each step, capture what is visible." and a 15-item enumeration of DOM things to gather (Elements, Components, Text, Links, ...) assume little intelligence and could be tightened into a few directive lines. | 3 / 5 |
Actionability | Mostly executable guidance: exact commands ("sites-preview start \"$PWD\"", "npm run dev -- --host 0.0.0.0 --port 4173 --strictPort"), concrete viewport ("390 x 844"), and specific Vite settings. Minor gaps remain, e.g. "generate a replacement with ImageGen" gives no invocation detail. | 4 / 5 |
Workflow Clarity | A 13-step numbered workflow with explicit validation checkpoints and feedback loops: step 3 page-correctness stop conditions, step 11 compare-and-fix against the original, and step 12's design-qa gate that must loop until "final result: passed" or report "blocked". | 5 / 5 |
Progressive Disclosure | References are one level deep and clearly signaled (critical-overrides, user-context, local-prototype-preflight, design-qa), but the ~15-line environment-specific "Previewing prototypes in ChatGPT Work Mode" section is inlined detail that belongs in local-prototype-preflight.md, and none of the referenced files ship inside this skill's bundle to verify. | 4 / 5 |
Total | 16 / 20 Passed |