The Agent Skills Directory

[SAFE]: The skill uses a local Python helper script (new_notebook.py) and bundled templates to generate notebook files, avoiding manual JSON construction and ensuring a consistent, safe structure.
[SAFE]: External dependencies are restricted to well-known, standard Python packages for data science (jupyterlab, ipykernel) and are managed through the standard tool uv.
[SAFE]: All file operations are scoped to the project environment, using logic to find the repository root and writing to a designated output directory.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to refactor existing notebooks and ingest user-provided content into new notebooks. 1. Ingestion points: User-provided notebook files for refactoring and notebook metadata/content inputs. 2. Boundary markers: None explicitly defined in the templates or script to isolate or ignore instructions embedded within processed notebook data. 3. Capability inventory: The skill allows the agent to generate code and documentation, and instructions suggest executing notebooks if the environment permits. 4. Sanitization: Titles are slugified for use in filenames, but cell content is otherwise populated based on user instructions without specific sanitization filters.

jupyter-notebook