image-media

Generate, edit, and process visual artifacts, diagrams, and multimedia assets. Use when creating UI mockups, visual explainers, image transformations, or media layouts (carousel, bento grid).

SKILL.md

GitHub
<!-- Generated from skills/image-media.md by tools/generate_integrations.py. Edit the source file, not this one. -->

Skill: Image & Media Handling

Purpose

Generate, edit, and process images and other media assets.

Tools Required

  • Image generation tools (text2im, image_gen)
  • Image editing tools (image_edit)
  • Video generation tools (video_gen)
  • Vision/analysis tools

General Principles

  • Only generate images when they add significant value to the response
  • If text alone is clear and sufficient, don't add images
  • Prefer built-in vision capabilities over OCR: OCR is high-cost, high-risk, last-resort
  • OCR libraries support English only

Image Generation

When to Generate Images

High value use cases:

  • Explaining processes visually
  • Browsing and inspiration
  • Exploratory context
  • Highlighting differences (before/after)
  • Quick visual grounding
  • Visual comprehension
  • Introducing people/places

Low value / avoid:

  • UI walkthroughs without exact current screenshots
  • Precise comparisons requiring accuracy
  • Speculation / spoilers / guesswork
  • Mathematical accuracy
  • Casual chit-chat / emotional support
  • Pure text-based tasks (definitions, grammar)
  • Writing / coding / data analysis

Layout Options

  • carousel (default): swipeable images in a row
  • bento: grid layout at top of response as cover; use for single entity deep-dives (person, place, sport team)

Image Parameters

  • Aspect ratio: 1:1 (default) or 16:9
  • Query: search terms to find relevant images
  • num_per_query: 1-5 images per query term
  • size: image dimensions
  • transparent_background: for PNG output

Image Editing

  • Modify existing images based on instructions
  • Add/remove elements
  • Alter colors, style transfer
  • Improve quality/resolution
  • Transform style (cartoon, oil painting, etc.)

Video Generation

  • Available when the tool supports it
  • Text-to-video with audio cues
  • Extending existing videos
  • Generating videos between specified first and last frames
  • Using reference images to guide content

Visual Analysis

  • Use vision capabilities to describe image content
  • Extract text from images when necessary
  • Analyze charts, diagrams, screenshots
  • Identify objects, people, scenes
  • Read UI layouts for automation

When to Use Multiple Image Groups

  • Long, multi-section answers: one image group per major section
  • Compare-and-contrast across categories
  • Timeline or era segmentation
  • Geographic or regional breakdowns
  • Ingredient → steps → finished result

Image Citations

  • When using web-sourced images, cite the source
  • For generated images, no citation needed
  • Don't fabricate image URLs