Auto-sync from website-creator

skills/image-edit/SKILL.md

@@ -0,0 +1,63 @@
+---
+name: image-edit
+description: Edit images using AI with text prompts and input images. Use this skill when the user wants to modify or transform an existing image with AI editing.
+---
+
+# Image Edit
+
+Edit images with AI by combining source images with text prompts via `python3 scripts/image_edit.py edit <prompt> <image_path> [options]`.
+
+## Commands
+
+| Command | Args | Description |
+|---------|------|-------------|
+| `edit` | `<prompt> <image_path> [--width W] [--height H] [--steps N] [--cfg-scale N]` | Edit image with prompt |
+
+## Options
+
+| Option | Default | Range | Description |
+|--------|---------|-------|-------------|
+| `--width` | 1024 | 128-2048 | Output image width in pixels |
+| `--height` | 1024 | 128-2048 | Output image height in pixels |
+| `--steps` | 40 | 5-100 | Number of inference steps |
+| `--seed` | null | 0-4294967295 | Random seed (null = random) |
+| `--cfg-scale` | 4 | 0-10 | True CFG scale for guidance |
+| `--negative-prompt` | "" | - | Negative prompt to avoid |
+
+## Examples
+
+```bash
+# Basic edit
+python3 scripts/image_edit.py edit "make it look like oil painting" photo.jpg
+
+# Style transfer
+python3 scripts/image_edit.py edit "convert to anime style" portrait.png
+
+# Object modification
+python3 scripts/image_edit.py edit "change the car color to red" street.jpg --steps 50
+
+# With negative prompt
+python3 scripts/image_edit.py edit "add a sunset background" landscape.png --negative-prompt "water, ocean"
+```
+
+## Workflow
+
+1. Provide a `prompt` describing the desired edit
+2. Provide an `image_path` to the source image (PNG, JPG, etc.)
+3. Script converts image to base64 and sends to API
+4. Saves edited image as `edited_[timestamp].jpg`
+5. Returns image path: `edited_1234567890.jpg [12345]`
+
+## Output Format
+
+- Success: `Image saved: filename.jpg [id]`
+- Error: `Error: message` (to stderr)
+- Images saved to current working directory as JPEG files
+
+## Notes
+
+- Requires `CHUTES_API_TOKEN` in environment
+- Supports up to 3 input images (currently uses first image)
+- Input file must be a valid image format (PNG, JPG, etc.)
+- Output is always JPEG format to save memory
+- Images are saved locally, not returned as base64 to save memory