The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses the uvx tool to download and run the marimo package from the public Python Package Index (PyPI). This is a standard procedure for running CLI utilities without permanent installation.
[COMMAND_EXECUTION]: The agent is instructed to execute shell commands, specifically marimo convert and marimo check, to process and validate notebook files on the local file system.
[REMOTE_CODE_EXECUTION]: Code from the marimo package is executed at runtime via the uvx runner to facilitate the file conversion process.
[PROMPT_INJECTION]: The skill processes untrusted Jupyter notebooks, which presents a surface for indirect prompt injection.
Ingestion points: Input .ipynb files are read and converted as described in SKILL.md.
Boundary markers: The instructions do not define specific delimiters to isolate untrusted notebook content during the review process.
Capability inventory: The skill includes file system access (read/write) and subprocess execution via the marimo CLI.
Sanitization: The skill relies on the conversion logic of the marimo library to correctly handle and parse the notebook structure.
[SAFE]: The skill references external documentation for the wigglystuff extension on koaning.github.io, which is a well-known and reputable source in the data science community.

jupyter-to-marimo