generate-image

Generate or edit an image based on the user's request.

Process

1. Parse the arguments:

   • The main text is the image prompt
   • If --edit is present, the next argument is the path to an input image for editing
   • Determine mode: "edit" if --edit flag is present, otherwise "generate"

2. Ask the user which provider to use with AskUserQuestion:

   • Gemini (gemini-3-pro-image-preview): Premium quality, professional asset production, up to 14 reference images
   • OpenAI (gpt-image-2): Best for text rendering and transparent backgrounds
   • xAI (grok-imagine-image-pro): Premium tier, higher quality, flat per-image pricing
   • All in parallel: Generate with all available providers and pick the best result

3. Ask the user where to save the output with AskUserQuestion:

   • Current directory (e.g., ./generated-image.png)
   • Custom path (let them type)

4. Execute the generation using tasks for progress tracking:

   If single provider selected: a. Create a task with TaskCreate:

   • subject: "Generate image with "
   • activeForm: "Generating image with ..." b. Mark it in_progress with TaskUpdate c. Run the script via Bash:

   bash "${CLAUDE_PLUGIN_ROOT}/scripts/<provider>.sh" \
     --mode <generate|edit> \
     --prompt "<prompt>" \
     --output "<output-path>" \
     [--input-image "<input-path>"]
   

   d. Mark the task completed (or note the error if it failed)

   If multiple providers selected: a. Create one task tracking the whole parallel run with TaskCreate:

   • subject: "Generate image with all providers in parallel"
   • activeForm: "Generating with Gemini, OpenAI, and xAI..." b. Mark it in_progress with TaskUpdate. c. Run scripts/run-all.sh in a single Bash call — it opens the streaming pane, forks all providers in parallel, captures their stdio under $DISPLAY_PANE_DIR/logs/, and closes the pane when every provider has returned:

   bash "${CLAUDE_PLUGIN_ROOT}/scripts/run-all.sh" \
     --mode <generate|edit> \
     --prompt "<prompt>" \
     --output-base "<base>" \
     [--input-image "<input-path>"]
   

   Each provider produces <base>-gemini.png, <base>-openai.png, <base>-xai.png. To run a subset, pass --providers gemini,openai. To pass per-provider tuning, use --gemini-extra "...", --openai-extra "...", --xai-extra "...". d. Mark the task completed when run-all.sh exits. Its exit status reports whether any provider failed; per-provider error details are in $DISPLAY_PANE_DIR/logs/<provider>.err and shown inline in the streaming pane as a red error banner.

5. After generation completes, confirm the output path(s) to the user. If multiple were generated, let the user know all files are available so they can compare.