# Run Python Code in a Sandbox

Users often need Python libraries that are unavailable in their local AI tool. This skill runs Python code through an external sandbox and returns text output or generated files.

## Install

```bash
npx skillstore add halt-catch-fire/python-executor
```

## Metadata

- Status: approved
- Slug: halt-catch-fire-python-executor
- Skillstore revision: r1
- Version status: missing
- Tree hash: e7dfc159ab8ec1467df6a41d7d9453591d31715785fa4a066460349610403e6f
- Author: halt-catch-fire
- GitHub username: halt-catch-fire
- License: MIT
- Repository: https://github.com/halt-catch-fire/skills/tree/main/tools/utilities/python-executor/
- Ref: a06681402992ceae98ba04d54cfd4ab004862696
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/halt-catch-fire-python-executor
- Manifest: https://skillstore.pages.dev/api/skills/halt-catch-fire-python-executor/manifest

## Capabilities

- Runs Python 3.10 snippets through the inference.sh belt CLI.
- Uses preinstalled libraries for data analysis, web scraping, media processing, documents, and 3D assets.
- Returns printed output from executed Python jobs.
- Returns generated files saved under outputs, including charts, CSVs, images, videos, PDFs, and 3D models.
- Supports configurable timeouts and a documented high-memory variant.
- Shows examples for HTTP requests, parsing, visualization, image processing, video creation, and API calls.

## Use Cases

- Analyze Data Tables: Run pandas and visualization tasks when a local Python environment is unavailable.
- Automate Developer Tasks: Execute short Python scripts for parsing, transformation, API checks, and repeatable utility work.
- Generate Media Files: Create images, videos, PDFs, and 3D files from scripted workflows.

## Prompt Templates

### Run a Basic Calculation

```
Use python-executor to run a short Python calculation and show the printed result. Keep the code simple and explain the output.
```

### Create a Chart

```
Use python-executor to analyze this table and create a bar chart saved under outputs. Return the summary and the generated file.
```

### Process Website Data

```
Use python-executor to fetch the approved public page, extract the main table, and return a clean CSV file under outputs.
```

### Build a Multi-Step Pipeline

```
Use python-executor to run a complete pipeline that cleans input data, creates a chart, exports a report, and uses a clear timeout.
```

## Limitations

- Requires the belt CLI and an inference.sh login or configured account.
- Code execution happens in an external sandbox, so sensitive data needs explicit approval.
- The documented environment is CPU-only and not designed for GPU workloads.
- Interactive displays are not supported; generated files must be saved under outputs.

## Best Practices

- Review code before running it, especially when it sends network requests or handles sensitive data.
- Save generated files under outputs so the executor can return them.
- Use clear timeouts and smaller test inputs before running large jobs.

## Anti Patterns

- Do not use it to run untrusted code without approval.
- Do not paste secrets, private data, or credentials into external execution jobs.
- Do not rely on interactive prompts, windows, or display calls for results.

## Security Audit

- Audited at: 2026-07-06T16:55:52.402\+00:00
- Summary: Many static backtick findings are Markdown formatting or library-name false positives. Real concerns remain because the skill authorizes the belt CLI, runs arbitrary Python in an external sandbox, and includes outbound HTTP examples. No prompt injection text was found in SKILL.md.

## Stats

- Views: 0
- Downloads: 6
- Favorites: 0
- Popularity score: 0