seer

Seer is a comprehensive visual automation and feedback skill designed for macOS environments. It provides agents with the ability to observe, record, and verify the state of GUI applications, making it an essential utility for automated UI testing, visual documentation, and interactive development workflows. By bridging the gap between agent logic and visual representation, Seer allows users to monitor application behavior in real-time or through historical snapshots, ensuring that visual feedback loops remain tight and reliable.

• Precise window-level screenshot capture using macOS native integration and process-based targeting.

• Video recording and frame extraction capabilities with customizable sampling strategies, scene detection, and summary generation.

• Full-display screen recording support using ffmpeg for long-running QA and verification tasks.

• UI mockup generation via JSON-based annotation specs, enabling dynamic creation of arrows, rectangles, spotlights, and descriptive text overlays.

• Natural language to Excalidraw wireframe conversion using a bundled UI component library for rapid prototyping and interface design.

• Visual regression testing framework that maintains baselines and generates diff reports to track UI changes over time.

• Seer requires Accessibility and Screen Recording permissions on macOS to interact with and capture application windows effectively.

• Output artifacts are organized by application and timestamp within a configurable directory (default: .seer/).

• Supports high-precision capture targeting specific process names or the current frontmost application.

• Ideal for developers and automation engineers performing repetitive verification tasks or creating descriptive visual bug reports.

• Use the loop_compare script to automate visual state verification against stored baselines for continuous monitoring.

• Leverage the provided Python utilities for advanced annotation and image comparison metrics.