interactive-notebook-demo
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to create notebooks that interact with external APIs and services, creating a surface where malicious instructions could be embedded in data processed by the notebook.
- Ingestion points: The references/cell-patterns.md file describes methods like store.get(key=...) and store.update(key=..., value=...) that ingest data from a hypothetical storage service.
- Boundary markers: No specific delimiters or 'ignore instructions' markers are required by the skill when interpolating external data into cells.
- Capability inventory: The generated notebooks execute code within a Jupyter kernel, which has the capability to perform network operations, file system access, and shell command execution.
- Sanitization: The skill does not mandate specific sanitization or escaping of external data before it is displayed or processed.
- [DYNAMIC_EXECUTION]: The skill provides patterns for generating Python scripts in percent-format (#%%) which are intended to be executed dynamically within an interactive kernel. It also includes a headless validation process that executes these scripts to ensure they run correctly.
- Evidence: The references/headless-validation.md file provides a bash command 'jupytext --to ipynb --execute --set-kernel demo ...' to execute the generated notebooks.
- [EXTERNAL_DOWNLOADS]: The validation script utilizes a package manager to fetch and run standard development tools from public registries.
- Evidence: The references/headless-validation.md file uses 'uv run --with jupytext --with nbclient --with ipykernel' to download and execute these packages from the Python Package Index.
Audit Metadata