icon-generator-skill

Agent skill for generating Android and iOS app icon sets from a single source image.

What it does

• Exports Android launcher icons across common mipmap-* sizes
• Exports iOS AppIcon.appiconset PNGs plus Contents.json
• Optionally exports Android adaptive icon foreground/background/XML assets
• Supports contain and cover fitting modes
• Supports custom padding color, extra square sizes, and rounded corners

Install With npx skills

This repository is structured as a root-level skill repository, so it can be installed directly with the open skills CLI:

npx skills add anhao/icon-generator-skill

Examples:

# Install globally for detected agents
npx skills add anhao/icon-generator-skill -g

# Install only for Codex
npx skills add anhao/icon-generator-skill -g -a codex

# Install only for Claude Code
npx skills add anhao/icon-generator-skill -g -a claude-code

# List available skills in this repo
npx skills add anhao/icon-generator-skill --list

The vercel-labs/skills CLI discovers skills from a repository root when that root contains SKILL.md, which is how this repository is organized.

Repository Layout

Main files:

• SKILL.md: skill trigger and workflow instructions
• scripts/icon_generator.py: bundled export script
• references/usage.md: argument and output reference

Dependency

python3 -m pip install 'Pillow>=10.0.0'

Direct Script Usage

Basic export:

python3 scripts/icon_generator.py input.png -o output_icons

Export with crop, rounded corners, and extra sizes:

python3 scripts/icon_generator.py input.png \
  -o output_icons \
  --mode cover \
  --background '#FFFFFFFF' \
  --extra-sizes 256 384 \
  --corner-radius 18

Export Android adaptive icons:

python3 scripts/icon_generator.py input.png \
  -o output_icons \
  --android-adaptive \
  --adaptive-background '#FFFFFF'

Manual Install

If you want to install it manually instead of using npx skills, place this repository in an agent skill directory and keep the skill folder name aligned with the skill trigger $icon-generator.

License

MIT