Content
78%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 compact, well-structured, largely executable skill body with a genuinely actionable step sequence. Its main defects are a minor aspect-ratio contradiction between steps 2 and 3 and the absence of a verification checkpoint for the negative-space requirement the deliverables rely on.
Suggestions
Resolve the aspect-ratio inconsistency: step 2 specifies 1.91:1 (matching the 1200×628 header size) while step 3 passes aspect_ratio=21:9 — pick one correct value for both places.
Include one complete copy-paste example of the muapi image generate invocation with placeholders substituted, so the executing agent doesn't have to assemble the command from fragments.
Add a quick post-generation check that the left half of the image is sufficiently uncluttered for title overlay before returning the title-placement guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — no concept explanations, just an inputs table, three numbered steps, deliverables, and brief notes. Minor trims exist: the description is duplicated verbatim as the bold header line, and the 'Leave empty space intentionally' note restates the 'ample negative space on the left side' already baked into the prompt template. This is anchor 4, not 5, because of those small redundancies. | 4 / 5 |
Actionability | Concrete guidance throughout: a specific CLI call ('muapi image generate (model=gpt-image-2-text-to-image, aspect_ratio=21:9)'), a templated prompt string, and a curl fallback for unaliased endpoints. It stops short of anchor 5 because no full copy-paste command example is given, and step 2's 'Aspect ratio: 1.91:1' contradicts step 3's aspect_ratio=21:9 (1200×628 ≈ 1.91:1, so one of the two values is wrong). | 4 / 5 |
Workflow Clarity | The sequence (derive metaphor → build prompt → call CLI → return deliverables) is clear and unambiguous, and as a simple one-shot generation it faces no destructive/batch validation cap. It misses anchor 5 because there is no checkpoint verifying the generated image actually contains the left-side negative space the title-placement deliverable depends on, and the 1.91:1 vs 21:9 contradiction leaves the executing agent uncertain which ratio to request. | 4 / 5 |
Progressive Disclosure | The skill is ~50 lines with no external references needed and no bundle files, and content is organized into well-labeled sections (Inputs, Steps, Deliverables, Notes, Trigger Keywords). Per the rubric's guideline for short single-purpose skills with no reference files, this earns a 5. | 5 / 5 |
Total | 17 / 20 Passed |