predictingthepast
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads large pre-trained model checkpoints (up to 16MB per chunk) from Google's Cloud Storage (
storage.googleapis.com/ithaca-resources/models). These downloads are necessary for the skill's primary function and target the vendor's own official infrastructure. - [DYNAMIC_EXECUTION]: The
scripts/run_inference.pyscript usespickle.load()to load model checkpoints and retrieval embeddings. Whilepickleis typically unsafe for untrusted data, in this context, it is used to load official vendor-provided model files which are validated against a hardcoded list of expected filenames (_VALID_MODEL_FILENAMES). - [COMMAND_EXECUTION]: The skill instructions prompt the user to run local Python scripts using
uv run. These scripts (preprocess.py,run_inference.py,visualize_results.py) perform text processing and model execution locally. The skill also starts a localhttp.serverto allow users to view the generated HTML results dashboard.
Audit Metadata