sandbox-script-executor

# Sandbox Script Executor This skill allows the agent to execute user-provided scripts within the secure sandbox environment and return their output. It supports multiple programming languages and includes mechanisms for automatic dependency management. ## Workflow When a user requests to execute a script or run code in the sandbox, follow these steps: ### 1. Obtain Script Details - **Script Content**: Ask the user to provide the full content of the script they wish to execute. - **Language**: Determine the programming language of the script (e.g., Python, Bash, Node.js). If not specified, ask the user. - **Dependencies (Optional)**: Inquire about any specific libraries or packages required by the script. If a `requirements.txt` (Python) or `package.json` (Node.js) is provided, use it. ### 2. Prepare Environment and Install Dependencies - **Save Script**: Save the user-provided script content to a file in the sandbox (e.g., `/home/ubuntu/script.py`, `/home/ubuntu/script.sh`, `/home/ubuntu/script.js`). - **Install Dependencies**: Before execution, attempt to install any specified dependencies. Refer to `references/execution_guide.md` for common dependency management commands (e.g., `pip3 install`, `npm install`). If dependencies are not explicitly mentioned, try to infer them or ask the user. ### 3. Execute Script - **Execute Command**: Use the appropriate command to execute the script based on its language. Refer to `references/execution_guide.md` for execution commands (e.g., `python3`, `bash`, `node`). - **Capture Output**: Ensure that both standard output (stdout) and standard error (stderr) are captured during execution. ### 4. Report Results - **Success**: If the script executes successfully, return the stdout to the user. - **Failure**: If the script fails, report the stderr to the user, along with any relevant error messages. Attempt to diagnose common issues (e.g., missing dependencies, syntax errors) and suggest solutions. ## Supported Languages and Dependency Management This skill supports: - **Python**: Execution via `python3`, dependency management via `pip3`. - **Bash**: Execution via `bash`, system package management via `apt-get` (requires `sudo`). - **Node.js**: Execution via `node`, dependency management via `npm` or `pnpm`. For detailed commands and examples, refer to `references/execution_guide.md`. ## Example Usage **User Request:** "Run this Python script: `print('Hello, Sandbox!')`" **Agent Action (Conceptual):** 1. Save script to `/home/ubuntu/temp_script.py`. 2. Execute `python3 /home/ubuntu/temp_script.py`. 3. Return output: "Hello, Sandbox!" **User Request:** "Execute this Node.js script that uses `axios`: `const axios = require('axios'); axios.get('https://api.example.com').then(res => console.log(res.data));`" **Agent Action (Conceptual):** 1. Save script to `/home/ubuntu/temp_script.js`. 2. Install `axios`: `npm install axios`. 3. Execute `node /home/ubuntu/temp_script.js`. 4. Return API response data.