The Agent Skills Directory

[No Code Issues] (SAFE): All included Python scripts are focused on physics and math (CFL limits, Fourier numbers, Von Neumann analysis). Code is clean and uses standard libraries.
[Data Exposure] (SAFE): No evidence of credential searching or access to sensitive paths (~/.ssh, etc.).
[External Downloads] (SAFE): Only standard numpy is required. No untrusted third-party packages or remote scripts.
[Indirect Prompt Injection] (LOW): The skill allows reading user-provided matrix files.
Ingestion points: matrix_condition.py (--matrix) and stiffness_detector.py (--jacobian).
Boundary markers: Absent.
Capability inventory: Bash, Write, Read.
Sanitization: Data is parsed into NumPy arrays; no executable logic is derived from input content.
[Prompt Injection] (SAFE): No instructions found that attempt to subvert the agent's core safety filters or extract system prompts.

numerical-stability