image-process

Image Process

When to Use

Use this skill when the user already has local image files and needs deterministic post-processing:

• Resize, crop, thumbnail, rotate, convert format, compress, set quality, or strip EXIF.
• Normalize a batch of images for web, social, app, marketplace, or document delivery.
• Produce non-destructive output files while preserving originals.

Do not use this skill for semantic AI edits such as "remove this person", "make the subject prettier", "change the background to a beach", or "generate a new image". Those require image design/editing or an AI image tool.

Workflow

1. Confirm inputs:

• exact file paths or glob range.
• target operation, dimensions, output format, output directory, and whether metadata should be stripped.

1. Protect originals:

• default to writing into an output directory or adding a suffix.
• do not overwrite originals unless the user explicitly confirms overwrite.

1. For batches:

• list or count matched files before processing.
• confirm if the operation is destructive, lossy, or large-scale.

1. Process in a stable order:

• auto-orient from EXIF, crop/fit, resize, convert, strip metadata, save quality.

1. Verify outputs:

• path, dimensions, format, file size, and failures.

Script

Use scripts/image_process.py when Python and Pillow are available.

Examples:

"$ORKAS_NODE" "$ORKAS_PC_DIR/bin/run-skill.cjs" image-process image_process -- --input cover.png --output-dir out --width 1080 --format webp --quality 85 --strip-exif
"$ORKAS_NODE" "$ORKAS_PC_DIR/bin/run-skill.cjs" image-process image_process -- --input "images/*.png" --output-dir out --width 1080 --height 1080 --fit cover --format jpg --quality 90

If the shared Skill Runner Python environment is missing Pillow, report the dependency instead of modifying the protected Marketplace install. An administrator may provision it in the shared runtime environment:

"$ORKAS_PYTHON" -m pip install Pillow

Do not install dependencies without user approval in restricted or shared environments.

Return Format

Return:

{
  "ok": true,
  "outputs": [
    {
      "path": "output.webp",
      "dimensions": "1080x1080",
      "format": "WEBP",
      "file_size": 123456,
      "operations": ["resize", "convert", "strip_exif"]
    }
  ],
  "failed": []
}

For failures, return the failed files and cause.

Dependencies

• Python 3.
• Pillow for image processing.
• Local read/write permissions.
• Optional HEIC/HEIF support may require extra libraries.

Limits and Known Issues

• This skill does not perform AI semantic edits.
• Format conversion can be lossy.
• EXIF stripping removes metadata such as camera, date, and GPS fields.
• Cropping with cover may remove image edges.
• Large batches should be previewed before execution.

Examples

User: "把 generated 目录下的 png 都转成 1080 宽的 webp，质量 85，去掉 EXIF。"

Handling: count matching files, confirm output directory, run the script with width, format, quality, and strip-exif, then return output paths and any failures.