generate-images
Generate Images (Media Renderer)
Reads IMAGE_SPEC.md, submits each entry's prompt to an AI image generation API, and saves the resulting PNG files to the project's images/ folder.
Output voice
Apply a lightweight human-voice pass to scope questions, review prompts, and
result reports. Keep prompts and user-provided Media Intent intact when
transporting them, and keep filenames, paths, commands, and state values exact.
Before approving any presentation-facing text, invoke the standalone unslop
Skill as a required full editorial pass using
DISCOVERY.json.editorialPreferences; preserve prompts and machine-readable
media metadata where their contract requires exact wording.
Protocol: resolve Media Scope, choose Generation Mode, review and report results, update only the owned media phase, leave it pending on cancellation or failure, and preserve unrelated phase records. Gemini setup remains local.