python-executor

What This Skill Does

The Python Executor is a robust, sandboxed environment designed for OpenClaw agents to execute arbitrary Python code. It acts as a powerful computational engine, allowing your AI agent to transition from simple conversation to executing complex logic, performing data crunching, and generating creative media assets. With over 100 pre-installed scientific and utility libraries, it functions as a cloud-based development environment that requires zero configuration on the local machine.

Installation

To integrate this skill, use the ClawKit CLI by running the command: clawhub install openclaw/skills/skills/okaris/python-executor. The installation verifies all dependencies and ensures the sandbox is configured for your specific architecture. Once installed, your agent automatically gains the ability to run scripts for data processing, web scraping, or file manipulation tasks. The underlying engine, inference.sh, ensures that every execution is isolated from your primary system, providing a secure bridge between your AI and your data.

Use Cases

This skill is highly versatile. It is ideal for data analysts who need to process CSV files or generate visualizations on the fly. Developers can leverage the browser automation stack—Playwright and Selenium—to scrape dynamic websites that block standard HTTP requests. For creative workflows, users can utilize OpenCV and MoviePy to manipulate video clips, add filters, or generate 3D models using Trimesh. It is also an essential tool for automating repetitive tasks, such as generating PDFs from data inputs or interacting with complex APIs.

Example Prompts

1. "Analyze this sales CSV data and generate a trend line chart using matplotlib. Save it to the output folder."
2. "Go to this URL, scrape the main article headlines, and return them as a bulleted list in the chat."
3. "Take the video file I uploaded, trim the first five seconds, and apply a grayscale filter using the Python executor."

Tips & Limitations

To maximize performance, be mindful of the 30-second default timeout; for resource-heavy video processing, ensure you utilize the high_memory flag. Because the sandbox is isolated, always explicitly save assets to the /outputs directory to ensure they persist and remain accessible for download. Avoid attempting to access local system resources, as the environment is strictly sandboxed for safety.