pymc-bayesian-modeling
Warn
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The file
references/advanced_workflows.mdcontains a code recipe for model persistence usingpickle.dump()andpickle.load(). Deserializing data using thepicklemodule is insecure and can lead to arbitrary code execution if the input file is maliciously crafted. While the skill recommends the NetCDF format for results, the inclusion ofpickleas a valid pattern for full model objects introduces a security risk for users loading external or third-party models. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided datasets, creating a surface for potential indirect prompt injection attacks.
- Ingestion points: External data enters the agent context via NumPy arrays and pandas DataFrames in
SKILL.md(Workflow Step 1) andreferences/advanced_workflows.md(Data Preparation). - Boundary markers: Absent; the instructions do not include specific delimiters or warnings to ignore instructions that might be embedded within the numeric or tabular data.
- Capability inventory: The skill includes file system read/write capabilities for saving and loading models (
to_netcdf,pickle). - Sanitization: The skill includes data standardization steps, but lacks specific security validation or filtering to prevent data-driven exploits.
Audit Metadata