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.
The body is a tight, well-structured, actionable recipe: concrete model IDs, exact prompts, parameterized inputs, and a curl fallback for endpoints without CLI aliases. Its main gaps are the absence of complete copy-paste command syntax and any error-recovery guidance for failed generations, which cap actionability and workflow clarity at 4.
Suggestions
Provide one complete example invocation per phase (e.g., `muapi image generate --model nano-banana --prompt "..." --aspect-ratio 9:16`) so commands are copy-paste ready.
Add a short error-recovery note: what to do if a `muapi` call or `muapi predict wait` poll fails or times out (retry, check key, fall back to curl).
Trim the duplicated purpose line at the top (it repeats the frontmatter description verbatim) to save tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence — an inputs table, two command-driven phases, and brief agent notes with no explanation of known concepts. Minor trimmable padding (the bolded first line repeats the description verbatim, and the long default Hindi script sits inline) keeps it below the level-5 "every token earns its place" anchor. | 4 / 5 |
Actionability | Concrete, specific guidance throughout — `muapi image generate` with model IDs (nano-banana, veo3.1-fast-image-to-video), exact image and video prompts, 9:16 aspect ratios, a curl fallback with polling. It falls short of level 5 because the full CLI invocation syntax is never spelled out as a copy-paste-ready command line. | 4 / 5 |
Workflow Clarity | A clearly sequenced two-phase process (generate image → present to user → animate after approval) with an explicit approval checkpoint between phases. Minor gap: no error-recovery guidance if a generation or poll fails, so it does not reach the level-5 validation/feedback-loop anchor. | 4 / 5 |
Progressive Disclosure | The body is roughly 50 lines for a single task with no need for external references (no bundle files exist), and it is organized into clean Inputs / Steps / Trigger Keywords / Notes sections — per the simple-skill scoring note, this earns the top score. | 5 / 5 |
Total | 17 / 20 Passed |