execute-code

name	execute-code
description	Execute code in isolated ACA Sandbox microVMs. Python (default), bash, JavaScript, TypeScript, and .NET. Use when the user wants to run code, process data, transform files, test scripts, or perform calculations.
metadata	{"author":"Troy Taylor","version":"1.0","pattern":"create-and-update"}

Execute Code

When to Activate

• "Run this code", "execute [code]", "process this data"
• "Calculate", "transform", "parse", "analyze"
• User provides code or asks for a script to be written and run

Workflow

1. Execute: execute_code(code, language) — auto-creates a session if needed
2. Upload files if user provides data: upload_file(session_id, file_name, content)
3. Review output: Check exit_code, stdout, stderr
4. Download artifacts if code generated files: download_artifact(session_id, file_path)
5. Save to memory if results should persist: save_memory(key, value)

Supported Languages

python (default), bash, javascript, typescript, dotnet

Pre-installed Python Libraries

All these are available without pip install. They are also auto-imported in Python sessions — you can use them directly (e.g., pd.DataFrame(...) works without import pandas as pd).

Document formats: python-docx, openpyxl, xlsxwriter, xlrd, python-pptx, pypdf, reportlab, extract-msg, olefile, mammoth, icalendar, vobject

Data & analysis: pandas (as pd), numpy (as np), matplotlib.pyplot (as plt), pillow (as PIL), pydantic

Productivity: jinja2, python-dateutil, qrcode, humanize, tabulate, cryptography

Web & parsing: requests, beautifulsoup4 (as bs4), lxml, feedparser, pyyaml (as yaml)

MCP & developer: mcp, httpx, fastapi, uvicorn, jsonschema

R: tidyverse, jsonlite, openxlsx (use language: "bash" with Rscript -e "...")

Output Format

Present results in a fenced code block. For errors, show stderr with suggestions.

Saving Files

To save generated files to OneDrive, use save_to_onedrive(session_id, file_path).