pymc-bayesian-modeling
Warn
Audited by Gen Agent Trust Hub on May 6, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The reference file
references/advanced_workflows.mdprovides code examples that use thepicklemodule to load data. - Evidence:
with open("model.pkl", "rb") as f: saved = pickle.load(f)in the 'Save/Load Patterns' section. - Risk: The
pickle.load()function is vulnerable to arbitrary code execution if it processes a file from an untrusted source. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data without using delimiters or instructions to ignore embedded commands.
- Ingestion points: Data entry points via NumPy arrays and Pandas DataFrames in
SKILL.md. - Boundary markers: Absent; no markers or instructions are provided to separate user data from agent instructions.
- Capability inventory: The skill enables the agent to execute complex Python code, install packages, and read/write to the filesystem.
- Sanitization: Absent; input data is used directly in modeling contexts.
- [EXTERNAL_DOWNLOADS]: The skill instructs the user and agent to install third-party libraries via a package manager.
- Evidence:
pip install pymc arviz numpy matplotlibin the 'Prerequisites' section ofSKILL.md. - [COMMAND_EXECUTION]: The skill provides numerous Python code snippets for the agent to execute in its environment to build and validate statistical models.
Audit Metadata