The Photograph skill enables the toolkit to capture visual snapshots of web pages and record screen activity. It uses Playwright for snapshots and FFmpeg for screen recording.
- High-Fidelity Captures: Uses a real Chromium browser to render pages exactly as a user would see them.
- Screen Recording: Capture video of the OS screen while a command is running.
- Dynamic Content Support: Automatically waits for network activity to settle during web captures.
- Process Sync: Recording automatically tracks the life cycle of a subprocess.
This skill requires playwright, chromium, and ffmpeg.
# Install playwright
pip install playwright
playwright install chromium
# Install FFmpeg (macOS)
brew install ffmpegTakes a standard 1280x800 screenshot and auto-generates a filename in data/screenshots/.
python3 .gemini/skills/photograph/scripts/take_screenshot.py "https://www.google.com"Captures the entire scrollable area of the page.
python3 .gemini/skills/photograph/scripts/take_screenshot.py "https://mlcommons.org/croissant" --full_pageRecord the screen while running a process (e.g., the Wizard).
python3 .gemini/skills/photograph/scripts/record_screen.py --command "python3 .gemini/skills/wizard/scripts/wizard.py https://youtu.be/dQw4w9WgXcQ 'My Demo'" --output data/recordings/wizard_run.mp4Record the screen for 5 seconds.
python3 .gemini/skills/photograph/scripts/record_screen.py --duration 5 --output data/recordings/quick_check.mp4The Orchestrator Expert can transition to the Photograph skill when:
- A website is highly dynamic and cannot be scraped by simple HTML parsers.
- A visual record of the data source is required for provenance.
- The user wants to record a video log of an automated process for auditing or demonstration purposes.