jupyter-notebook
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill workflow involves executing a bundled Python script (
scripts/new_notebook.py) usinguv run. This script is used locally to scaffold notebook files from templates, performing standard filesystem operations like reading assets and writing output files. - [INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided titles and objectives to populate the generated notebooks.
- Ingestion points: User input for the notebook title is passed to the scaffolding script in
SKILL.mdvia command-line arguments. - Boundary markers: User-provided titles are wrapped in Markdown headers (e.g., '# Experiment: TITLE') within the notebook's JSON structure.
- Capability inventory: The skill can create and write
.ipynbfiles to the local filesystem. - Sanitization: Filenames are generated using a slugify function, and notebook content is safely serialized using standard JSON formatting.
Audit Metadata