Omniscient Execution System

Problem

Many automation tasks are not single-capability jobs: they require reasoning, script generation, tool calls, system control, and sometimes window, hardware, or IoT device operations. Using prompts, scripts, or CLI steps separately often means manually stitching context and repeatedly checking outputs, so complex tasks can break into disconnected steps. Omniscient treats these actions as composable primitives in one execution chain.

How It Works

It organizes tasks around three entry points:

• Cognition layer: mode-direct, improvement, transfer, and construction modes handle analysis, copywriting, solution design, and open-ended exploration.
• Execution layer: the model generates shell commands or Python scripts, writes them to files, runs them, captures output, and repairs failures up to two times.
• Control layer: modules such as window_manager.py, process_manager.py, hardware_controller.py, and iot_controller.py manage windows, processes, volume, network, GPU, storage, Bluetooth, and Home Assistant.

The orchestration engine combines five primitive types: script commands, code execution, tool calls, skill plugins, and cognitive output. It supports conditional branches, loops, parallel steps, checkpoints, and cognitive checks between steps, such as analyzing prior output before choosing the next command.

Boundaries

The skill is oriented toward Windows desktop environments, and Python paths, virtual environments, and script directories should match the documented layout. Some modules auto-install dependencies such as psutil, pyserial, bleak, and opencv-python-headless. Destructive operations like shutdown, process termination, and network adapter disabling require user confirmation. GPU details, detailed temperatures, and BLE features depend on external tools such as nvidia-smi and OpenHardwareMonitor.