fluidsim

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [Indirect Prompt Injection] (MEDIUM): Ingestion of external simulation data presents a vulnerability surface. 1. Ingestion points: load_sim_for_plot, load_state_phys_file, and os.listdir in output_analysis.md and advanced_features.md. 2. Boundary markers: Absent. 3. Capability inventory: Cluster job submission (cluster.submit_script), MPI subprocess execution (mpirun), and file system write operations. 4. Sanitization: Absent.
  • [Dynamic Execution] (MEDIUM): The skill demonstrates generating Python scripts at runtime for submission to compute clusters using f-string templates in advanced_features.md. This pattern is susceptible to code injection if variable components (e.g., parameter names or values) are derived from untrusted agent inputs.
  • [Command Execution] (LOW): The documentation includes instructions for executing parallel simulations via mpirun, involving standard subprocess creation and management.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 01:35 PM