hand-tracking-toolkit
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone the
hand_tracking_toolkitfrom the official Facebook Research (Meta) GitHub repository and install dependencies from a requirements file. This is a reference to a well-known technology organization. - [DYNAMIC_EXECUTION]: The usage examples utilize
pickle.load()to process prediction data. Thepicklemodule is known to be insecure against erroneous or maliciously constructed data, as it can be exploited to execute arbitrary code during deserialization. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which creates a potential attack surface.
- Ingestion points: Data enters the context via
predictions.pklfiles and ground truth sequence directories. - Boundary markers: No delimiters or instructions to ignore embedded commands are present in the processing logic.
- Capability inventory: The skill executes Python scripts (
evaluate.py,visualize.py) and utilizes several libraries for file I/O and reporting. - Sanitization: There is no evidence of validation or sanitization for the content of the loaded pickle files or dataset annotations.
Audit Metadata