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 well-structured, actionable two-phase workflow with concrete commands, models, and ready-to-use prompts, and it appropriately keeps everything inline given its small size. The main improvement opportunities are minor: de-duplicate the repeated description line, make the muapi invocations literal command lines, and add a light validation step for generated images.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body (~48 lines) is mostly lean: a compact inputs table, two phase sections whose long prompt strings are the actual payloads (earning their tokens), and short execution notes. Minor trimmable padding exists — the description is repeated verbatim as a bold subheading, and the generic curl fallback note partially re-explains CLI mechanics Claude could infer. Not 5: those small redundancies; not 3: there are no multi-sentence explanations of concepts Claude already knows. | 4 / 5 |
Actionability | Concrete guidance throughout: named commands with model IDs ('muapi image generate (model=gpt-image-2-text-to-image)', 'muapi image edit (model=nano-banana-2-edit)'), full copy-ready prompt text, aspect ratios, and placeholder substitution rules. Not 5: the command invocations are structured as labeled bullets rather than literal, complete command lines (flags/argument syntax unspecified), and the curl fallback is a template rather than a filled example; not 3: the guidance is far beyond pseudocode — an agent can execute it with minor gap-filling. | 4 / 5 |
Workflow Clarity | The two-phase sequence is explicit and conditional (Phase A runs only 'If {{empty_room_image}} is not provided'; Phase B consumes Phase A's output or the user image), with a user-checkpoint between phases ('Present the empty room to the user') and a next-step suggestion. Not 5: there are no validation steps (e.g., verifying the generated base room is actually empty or that the edit succeeded) — though the guideline's destructive/batch cap does not apply to image generation; not 3: the sequence, conditionals, and data flow are all clearly defined, with only minor validation gaps. | 4 / 5 |
Progressive Disclosure | This is a simple skill under 50 lines with no bundle files (references/, scripts/, assets/ are absent) and no need for external references; per the rubric's simple-skill guidance, well-organized sections alone warrant a 5. Sections (Inputs, Steps with phases, Trigger Keywords, Notes) are clearly headed, and everything is appropriately inline at one level with no nested or buried references. | 5 / 5 |
Total | 17 / 20 Passed |