Content
87%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is concise, highly actionable, and well-structured for a single-purpose recipe, with concrete models, commands, and prompt templates. Its main weakness is the absence of an explicit error-recovery or validation loop for the generation and polling operations.
Suggestions
Add a short feedback loop for failed or low-quality generations (e.g., check muapi predict result status; if failed, retry with adjusted prompt before proceeding) to lift workflow clarity above 2.
State the expected output of each phase explicitly (e.g., an image artifact in Phase A, two video clips in Phase B) so validation of each step's result is unambiguous.
Clarify the ffmpeg concat step from the note into an explicit numbered step with the actual command, since merging the shots is part of the final deliverable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and efficient — a compact inputs table, specific numbered prompts, and brief notes — and it assumes Claude's competence without explaining what models or cinematography are, so most tokens earn their place. | 3 / 3 |
Actionability | It provides concrete, executable guidance: named models (e.g. 'nano-banana-2-edit', 'grok-imagine-image-to-video'), specific CLI commands ('muapi image generate', 'muapi video from-image'), full prompt templates with placeholders, and a copy-paste-ready curl fallback with polling, making it fully actionable. | 3 / 3 |
Workflow Clarity | The two-phase sequence is clearly numbered and includes an explicit approval checkpoint before Phase B, but for generation/polling operations there is no validation or feedback loop describing what to do if generation fails or quality is poor, capping workflow clarity at 2 per the batch/risky-operation rule. | 2 / 3 |
Progressive Disclosure | The skill is under 50 lines with no need for external bundle files, and it is well-organized into clear sections (Inputs, Steps, Trigger Keywords, Notes), so per the simple-skill scoring note progressive disclosure earns a 3. | 3 / 3 |
Total | 11 / 12 Passed |