Content
72%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 well-structured skill body: dense reference tables, executable commands, complete markup snippets, and an exemplary load-on-demand reference structure. The main weakness is the generation pipeline's lack of validation checkpoints — especially for the batch mode that makes multiple paid API calls — plus placeholder-styled model routing rows and a few vacuous cost-tracking lines.
Suggestions
Add a verification step to the Generation Pipeline (e.g., confirm the output file exists with the expected dimensions/aspect ratio before running the post-generation checklist), and for batch mode, validate the first variation before spending on the remaining N-1.
Replace the placeholder Model Routing entries ("Installed MCP/tool default @ 1K") with the concrete model identifiers from references/gemini-models.md, and show the actual `set_aspect_ratio` invocation syntax alongside the pipeline step.
Delete or concretize the "Approximate costs" stub (either state the numbers with a dated caveat or link directly to references/cost-tracking.md), and replace the empty presets code comment with a real command or a pointer to references/presets.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is table- and command-driven with almost no concept explanation Claude already knows. It loses a point for vacuous filler like "Approximate costs: Verify current pricing in the installed MCP/tool configuration before quoting" and a presets section whose code block contains only a comment ("# Use the installed Banana MCP/tool configuration to list presets") rather than instructions. | 4 / 5 |
Actionability | Mostly executable guidance: `./extensions/banana/install.sh`, `magick output.png -quality 85 output.webp`, complete ImageObject JSON and og: meta tag snippets, plus parameterized use-case tables. Gaps keep it below 5: the Model Routing table uses unactionable placeholders ("Installed MCP/tool default @ 1K") and "Set aspect ratio via `set_aspect_ratio` MCP tool" is given without its parameters. | 4 / 5 |
Workflow Clarity | The 6-step pipeline (identify use case, apply defaults, set aspect ratio, construct brief, generate, post-gen checklist) is clearly sequenced and paired with an error-handling table, but there are no validation checkpoints on the output. The batch command generates N paid API variations with no verification step before or after spending, so the batch-operations cap applies — a 4 would require explicit checkpoints such as confirming the generated file exists with expected dimensions. | 3 / 5 |
Progressive Disclosure | All nine referenced files (prompt-engineering.md, prompt-templates.md, prompt-adaptation-safety.md, gemini-models.md, mcp-tools.md, post-processing.md, cost-tracking.md, presets.md, seo-image-presets.md) exist in the bundle, are one level deep, and are listed under "Reference Documentation" with one-line descriptions and explicit "Load on-demand. Do NOT load all at startup" guidance. This matches the anchor-5 pattern of a clear overview with well-signaled references. | 5 / 5 |
Total | 16 / 20 Passed |