NYC

fluidsim

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS] (LOW): The skill instructs users to install fluidsim and related dependencies like fluidfft and mpi4py from PyPI using the uv package manager.- [COMMAND_EXECUTION] (MEDIUM): The skill provides patterns for generating Python scripts at runtime and executing them using mpirun or cluster-specific submission tools (e.g., Calcul8). This dynamic script generation and subsequent system-level execution is a core functionality but poses a significant risk if simulation parameters are not strictly validated.- [PROMPT_INJECTION] (LOW): The skill identifies a surface for indirect prompt injection when loading data from external simulation directories or HDF5 files. Ingestion points: Functions like load_sim_for_plot and h5py.File allow loading of external data. Boundary markers: None identified in the provided documentation. Capability inventory: Includes file-writing, system command execution (MPI), and cluster job submission. Sanitization: No sanitization or validation logic is presented for file paths or the contents of simulation files.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 05:58 PM