Skip to main content
2026/8/13 23:17 (UTC+8) · Docs Update 📚 Two advanced image pages: workflow orchestration and a prompt doctor skill Consumer image products beat a bare API call on the same model, and the difference is not in the model weights — it is the layer wrapped around the model. Two new pages take that layer apart and hand you a skill you can install directly:
  • Advanced Image Generation: Workflow and Realism: a five-step pipeline of rewrite layer, reference anchoring, client-side parallel sampling, vision-model selection and step-by-step retouching, plus a copy-paste realism vocabulary, three complete example prompts and a three-tier cost table.
  • Image Prompt Doctor Skill: a two-file Agent Skill that scores a prompt against six elements and returns a rewritten version before you generate, and can read a finished image back to point out which sentence was not executed; defaults to gpt-5.6-luna, pure Python standard library, nothing to install.
In our test, a casual prompt rewritten by the diagnosis and regenerated as-is lost every invented brand name and stray prop.
← Back to Live Updates · 📚 Monthly Archive