Agent-Skills.md

Agent Skills: Google Image Generation Skill

>

UncategorizedID: ypfaff/google-image-gen-skill/google-image-gen

Repository

ypfaff/google-image-gen-skill
ypfaffLicense: MIT
3

Install this agent skill to your local

pnpm dlx add-skill https://github.com/ypfaff/google-image-gen-plugin/tree/HEAD/skills/google-image-gen

Skill Files

Browse the full folder contents for google-image-gen.

Download Skill

Loading file tree…

skills/google-image-gen/SKILL.md

Skill Metadata

Name
google-image-gen
Description
>

Google Image Generation Skill

Generate images from text prompts using Google's Gemini API.

First-Time Setup (Once Per Context)

Run these commands once at the start of a session:

${CLAUDE_PLUGIN_ROOT}/scripts/check_env.sh
cd ${CLAUDE_PLUGIN_ROOT} && uv sync && cd -

If the environment check fails, the user needs to configure their API key. Recommended: Create ~/.config/google-image-gen/.env with:

GOOGLE_AI_API_KEY=your_key_here

Get an API key from https://aistudio.google.com/apikey

Alternative: Export as environment variable: export GOOGLE_AI_API_KEY=your_key

Usage

Important: The plugin runs from its own directory. Use --cwd to ensure output files land in the user's project, not the plugin cache.

ORIG_CWD="$(pwd)" && cd ${CLAUDE_PLUGIN_ROOT} && uv run python main.py --cwd "$ORIG_CWD" <output_path> "<prompt>" [options]

Note: When the user specifies a path like output.png or images/photo.png, pass it as-is — the --cwd parameter ensures it resolves relative to the user's project root.

Options

| Option | Short | Description | |------------|-------|-------------------------------------------------------------------------| | --style | -s | Style template (.md file with {subject} placeholder) | | --ref | -r | Reference image for style (repeatable, max 14) | | --edit | -e | Edit existing image instead of generating | | --aspect | -a | Aspect ratio: 1:1, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 |

Examples

Simple Generation

ORIG_CWD="$(pwd)" && cd ${CLAUDE_PLUGIN_ROOT} && uv run python main.py --cwd "$ORIG_CWD" output.png "A red apple on a wooden table"

Save to subdirectory

ORIG_CWD="$(pwd)" && cd ${CLAUDE_PLUGIN_ROOT} && uv run python main.py --cwd "$ORIG_CWD" images/output.png "A red apple"

With Aspect Ratio

ORIG_CWD="$(pwd)" && cd ${CLAUDE_PLUGIN_ROOT} && uv run python main.py --cwd "$ORIG_CWD" thumb.png "Mountain landscape" --aspect 16:9

Edit Existing Image

ORIG_CWD="$(pwd)" && cd ${CLAUDE_PLUGIN_ROOT} && uv run python main.py --cwd "$ORIG_CWD" output.png "Change the sky to sunset" --edit input.png

With Reference Image

ORIG_CWD="$(pwd)" && cd ${CLAUDE_PLUGIN_ROOT} && uv run python main.py --cwd "$ORIG_CWD" output.png "Same style but with a car" --ref reference.png

Multiple Variations

Generates numbered outputs (output_1.png, output_2.png, etc.):

ORIG_CWD="$(pwd)" && cd ${CLAUDE_PLUGIN_ROOT} && uv run python main.py --cwd "$ORIG_CWD" output.png "cat" "dog" "bird"

Workflow

  1. First use in context: Run setup commands (check_env.sh + uv sync)
  2. Generate images as needed
  3. Report output paths to user

Notes

  • Paid API tier recommended (free tier has strict rate limits)
  • Output directories are created automatically
  • Default aspect ratio is 16:9