Photo Edit Brief

Find out whether the photo edit you want can survive a re-render at all - before you spend anything. There is no image-in/image-out model here: the renderer never sees your photograph, it reads words and paints something new. So eight edit types are rated against what description can actually carry (three survive, three degrade, two cannot work), every "keep this unchanged" in your request is translated into words a blind renderer can act on, and a thirteen-line linter blocks the four defects that would waste a paid render. Then the writer reads your photo and drafts the brief, GPT Image renders it, and a third lane compares the result against your original and reports what actually survived. Taxonomy, prompt schema and per-slug prompting rules from the imagegen skill by @openai (github.com/openai/skills).

Every public app is built from a security-scanned skill and must pass a clean scan — skill and frontend — before it can be listed. Have a skill of your own? Turn it into an app — or read the step-by-step walkthrough.