antigravity-image
// Generate images using the internal Antigravity Sandbox API (Gemini 3 Pro Image). Supports text-to-image generation via internal Google endpoints.
$ git log --oneline --stat
stars:1,933
forks:367
updated:March 4, 2026
SKILL.mdreadonly
SKILL.md Frontmatter
nameantigravity-image
descriptionGenerate images using the internal Antigravity Sandbox API (Gemini 3 Pro Image). Supports text-to-image generation via internal Google endpoints.
metadata[object Object]
Antigravity Image Generation
Use the internal Google Antigravity sandbox to generate images with the gemini-3-pro-image model. This skill bypasses standard API restrictions by impersonating the VS Code plugin environment.
Usage
Generate an image from a prompt:
python3 {baseDir}/scripts/generate_image_antigravity.py --prompt "A futuristic cityscape with flying cars" --filename "city.png"
How it works
- Authentication: Automatically retrieves OAuth tokens and Project ID from the local
auth-profiles.json. - Impersonation: Uses the specific
antigravityUser-Agent and internal sandbox headers. - Reliability: Includes automatic retry logic for
503 Service Unavailableerrors.
Parameters
--prompt/-p: Text description of the image to generate.--filename/-f: Path where the resulting PNG will be saved.
Notes
- The script prints a
MEDIA:path which OpenClaw uses to automatically upload the image to the chat channel. - Requires an active Antigravity session/login in the environment.
- See internal-api.md for technical details on the impersonated endpoint.