vision-click - Vision-based coordinate click using Codex CLI
Clicks non-DOM elements via screenshot analysis using Codex CLI and mouse coordinates.
Tags
Updated: 2026-09-17Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Capture browser screenshot
- Analyze screenshot with Codex CLI
- Correct DPR coordinates
- Execute mouse click on coordinates
- Verify target before click
- Resize viewport or clip region
Inputs
- Target element description
- Browser screenshot
- CDP port and browser script path
- Optional viewport and clip flags
- Optional observation bundle JSON
Outputs
- Mouse click event on browser
- Command exit code
Requirements
- agbrowse running Chrome
- Codex CLI installed
