image-editing · Image Processing
General image processing: deterministic pixel-level operations on existing images — resize, crop, pad to platform aspect ratios, format conversion (png/jpg/webp), compress to a target size, add text or image watermarks, rounded corners, multi-image collage, thumbnails, and image metadata reading.
All operations run through deterministic Pillow scripts. Built-in platform size recommendations for social publishing (e.g. Xiaohongshu vertical 1080×1440): pad to preserve the frame, or center-crop for edge-to-edge. Multi-step operations can be chained.
Example invocation: "Pad this image to 1:1 and add a bottom-right watermark."
Full brief
Positioning
image-editing is a "workshop" for existing images: pixel-level processing only — no visual design, no content generation. For designed-from-scratch graphics use the card-* series; for AI-painted images use ai-image-gen.
Core capabilities
- Resize: by width/height/percentage, with aspect-preserving option.
- Crop: coordinate crop or center crop to given dimensions.
- Pad to ratio: pad any image to 1:1 / 9:16 / 16:9 and other ratios (the core of social-size adaptation, custom background colors).
- Format conversion: png/jpg/webp with quality control.
- Compression: fixed quality, or approach a target file-size cap (jpg/webp).
- Watermark: text (position/opacity/size/color, auto-finds CJK fonts) or image watermark.
- Rounded corners: transparent PNG output; 50% radius gives a circular avatar.
- Collage: horizontal / vertical / N×M grid, uniform cell sizes and gaps.
- Thumbnails: by longest-edge pixels.
- Image info: JSON output (dimensions/format/ratio/size/alpha channel).
Workflow
- Confirm input images exist; clarify target operations and parameters
- Look up platform size recommendations when adapting for social
- Run the corresponding subcommands (chainable, e.g. resize then compress)
- Report output paths, dimensions, and sizes
Inputs & outputs
| Input | Notes |
|---|---|
| Image files | Required — one or more existing image paths (Pillow-supported formats) |
| Target operations & params | Required — dimensions, ratios, quality, watermark text, collage mode, etc. |
Output: processed image files (user-specified path or default outputs/<theme>/); each command echoes path, dimensions, and size.
Boundaries with adjacent skills
| Skill | Lane |
|---|---|
| image-editing (this) | Processes existing images; pixel-level only |
| card-* series | HTML design → rendered new graphics |
| ai-image-gen | AI-generated new imagery |
| chart/infographic | Charts and infographics generated from data/text |
Fit
- Resizing images for social publishing (Xiaohongshu 3:4, Moments 1:1, video covers 9:16)
- Compressing images under platform upload limits
- Batch watermarking, thumbnail generation, multi-image grids
Before you start
- Pillow dependency (installed); purely local processing, no third-party cost.
- Built-in boundary checks: missing files, out-of-bounds crop boxes, or invalid ratios/colors exit with an error.
image-editing is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.