Set AI image generation house rules for Midjourney and DALL-E

You are an art director. AI-generated images drift wildly across operators. Build me a set of house rules so a contractor using Midjourney, DALL-E or similar produces on-brand images consistently.

Deliver: (1) the style description in one paragraph (lighting, palette, composition, treatment), (2) the prompt template (a fillable structure with anchor language we always include), (3) the negative prompt list (what we always exclude - 'no glossy 3D render, no neon glow, no centred composition'), (4) 5 example prompts for typical asset types (founder portrait, blog header, social tile, product mockup, abstract concept), (5) the model and parameter defaults (Midjourney version, aspect ratio, stylize value, weird value), (6) the post-generation edit rules (Photoshop / Photopea steps - colour cast correction, grain, crop), (7) the 'reject and regenerate' triggers (when to stop tweaking and re-prompt), (8) the never-publish list - signs of AI-tell that disqualify an image (warped hands, inconsistent text, generic stock-photo lighting), (9) the human-in-the-loop rule - when AI is the starting point vs the final asset. Plain English.

Brand: {{brand}}
Visual style description: {{style}}
Hero references (real images we want to echo): {{references}}
Things the AI keeps doing wrong: {{ai_problems}}