The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes complex shell scripts (Allrun, Allmesh) and SLURM batch scripts at runtime. These scripts are executed via the run_case and run_openfoam_command tools and incorporate user-influenced parameters and file names.
[REMOTE_CODE_EXECUTION]: Uses the run_python_script MCP tool to execute dynamically generated Python code (utilizing PyVista) for visualizing simulation results. This allows the execution of code generated by the agent within its runtime environment.
[DATA_EXFILTRATION]: The instructions explicitly direct the agent to use 'local file tools' to read repository configuration (config/user.yml) and copy STL files into the simulation runs tree (e.g., ~/foamagent/runs/). This encourages the agent to bypass the restricted file I/O tools provided by the MCP server (read_case_file/write_case_file) and access the host's filesystem directly.
[PROMPT_INJECTION]: The skill possesses a high surface area for Indirect Prompt Injection (Category 8). It ingests untrusted user requirements and processes external solver logs to diagnose errors, which then influences the generation of executable scripts and shell commands without explicit sanitization or boundary markers.

foam