What problem does it solve? Raw prompts for AI image and video generation are often vague, contradictory, or mismatched to the target provider, producing results that miss the intended subject, style, or composition. This Skill rewrites rough prompts into specific, provider-aware prompts using a structured critique loop instead of blind rewriting. ## Core Features & Use Cases - Optimize-Evaluate-Iterate Loop: Rewrites a prompt, scores it against the real requirement (subject clarity, context, contradictions, abstract terms), then fixes weaknesses until no major issues remain. - Provider-Aware Tuning: Adapts prompt structure for Flow/Veo (one camera move plus one action), ChatGPT/DALL-E (natural descriptive sentences), Gemini (layout and multi-subject editing), and Grok (concise phrasing). - Typed Placeholder Extraction: Pulls reusable variables like {subject}, {ratio}, {style}, and {color} with suggested values so prompts can be templated and reused. - Failure Diagnosis: When a generated image is wrong, diagnoses the cause (unanchored subject, missing style block, baked-in text, drifting character) and maps it to a concrete prompt fix. - Use Case: You generated a product photo and the style drifted from your reference. Paste the original prompt and the bad result; the Skill returns a corrected prompt, a diagnosis of what went wrong, and a before/after comparison. ## Quick Start Ask the assistant to optimize your rough image prompt for a specific provider, for example: optimize this prompt for Veo — a cat sitting on a table, cinematic style.
