computer▌

../../../engineering-team/senior-computer-vision/SKILL.md

AI agents that perceive screens, reason about actions, and control computers like humans do. \n \n Implements the perception-reasoning-action loop: capture screenshot, analyze with vision-language model, execute mouse/keyboard operations, repeat \n Covers Anthropic's Computer Use (Claude 3.5 Sonnet and Opus 4.5), with tool support for screenshots, mouse/keyboard control, bash execution, and file editing \n Requires sandboxed environments (Docker containers with virtual desktops) to isolate agent

Vision-driven desktop automation for native apps using natural language commands and screenshots. \n \n Controls macOS, Windows, and Linux desktops entirely from visual input; no DOM or accessibility labels required \n Operates synchronously with a screenshot-analyze-act loop: connect, observe screen state, execute high-level actions via natural language prompts, then disconnect \n Requires a vision-capable AI model (Gemini, Qwen, Doubao, or similar) configured via environment variables; support

Computer vision enables machines to understand visual information from images and videos, powering applications like autonomous driving, medical imaging, and surveillance.

Analyze events through the disciplinary lens of computer science, applying computational theory (complexity, computability, information theory), algorithmic thinking, systems design principles, software engineering practices, and security frameworks to evaluate technical feasibility, assess scalability, understand computational limits, design efficient solutions, and identify systemic risks in computing systems.

Role: Advanced Vision Systems Architect & Spatial Intelligence Expert

The fundamental architecture of computer use agents: observe screen, reason about next action, execute action, repeat. This loop integrates vision models with action execution through an iterative pipeline.

Production-grade computer vision expertise for image/video processing, object detection, model training, and inference optimization. \n \n Covers advanced architectures, real-time inference systems, and scalable data pipelines with distributed computing frameworks (Spark, Airflow, Kafka) \n Includes object detection optimization, model deployment strategies, and production monitoring with MLflow and Weights & Biases \n Supports PyTorch, TensorFlow, YOLO, and vision transformers with contain

Expert guidance for computer vision development using OpenCV, PyTorch, and deep learning techniques. \n \n Covers traditional image processing (filtering, edge detection, morphological operations, geometric transformations) and modern deep learning approaches (YOLO, Faster R-CNN, transfer learning with pre-trained models) \n Includes feature detection and matching (SIFT, ORB, FLANN), object detection with proper bounding box handling, and video processing with frame-by-frame pipelines and object

Gemini 2.5 Computer Use browser automation with Playwright-based agent loops and safety confirmations. \n \n Implements a screenshot-to-action cycle: capture screen, send to Gemini, parse function calls, execute in Playwright, return results until task completion or turn limit \n Supports multiple browser options: bundled Chromium (default), Chrome/Edge channels via COMPUTER_USE_BROWSER_CHANNEL , or custom executables like Brave \n Includes safety confirmation workflow that prompts users before