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 an efficient, actionable single-purpose skill: concrete model IDs, a prompt template, input handling, and a fallback path for models without CLI aliases. Main gaps are minor — a complete literal CLI invocation example, a success-check step before presenting the video, and small duplicated text.
Suggestions
Show one complete executable command, e.g. `muapi video generate --model veo3.1-text-to-video --prompt "..." --aspect-ratio {{aspect_ratio}} --duration {{duration}}`, so the flag syntax is copy-paste ready.
Add an explicit status step (e.g. `muapi predict wait <request_id>` and only present the video on success) before the 'present the video' step.
Remove the duplicated tagline under the title and the 'Emphasize "no cuts, no transitions"' note, since the prompt template already encodes that.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — an inputs table, one generation step, short notes, and agent notes — with no explanation of concepts Claude already knows. Minor trimming is possible: the tagline duplicates the frontmatter description verbatim, and the note 'Emphasize "no cuts, no transitions" in the prompt' repeats guidance already baked into the prompt template. This fits 'minor instances of over-explanation that could be trimmed' rather than 5's 'every token earns its place'. | 4 / 5 |
Actionability | Concrete guidance throughout: named CLI commands with model IDs ("muapi video generate (model=veo3.1-image-to-video)"), a literal prompt template with {{placeholders}}, an auth precondition, and a fallback curl command with polling. Not 5 because the exact full command line with its flags (how prompt/aspect_ratio/duration are passed to `muapi video generate`) is never shown as a single executable invocation — the sub-bullets imply but do not spell out the syntax. | 4 / 5 |
Workflow Clarity | The sequence is clear and unambiguous for a simple skill: gather inputs, configure auth if unset, pick the model based on whether reference_image is provided, generate, then present with follow-up suggestions. Not 5 because there is no checkpoint between submission and presenting (e.g. confirming the prediction succeeded via `muapi predict wait` in the main path, not only the fallback), leaving a minor validation gap. | 4 / 5 |
Progressive Disclosure | The skill is under 50 lines, needs no external references, and none exist (references/, scripts/, assets/ are absent and the body points to no other files). Sections are well organized (Inputs, Steps, Notes, Trigger Keywords, agent notes), which per the rubric's simple-skill guideline earns a 5 with just well-organized sections. | 5 / 5 |
Total | 17 / 20 Passed |