The Agent Skills Directory

[SAFE]: The skill implements strong security practices across all analyzed components. Specifically:
[DATA_EXPOSURE]: File access is limited to reading matrix data in .npy or text formats. Scripts like scripts/scaling_equilibration.py and scripts/sparsity_stats.py include explicit file size limits (500 MB) and dimension checks (100,000) to prevent resource exhaustion attacks.
[REMOTE_CODE_EXECUTION]: All matrix loading via NumPy uses allow_pickle=False, which is a critical security measure to prevent arbitrary code execution from malicious data files.
[COMMAND_EXECUTION]: The skill's allowed-tools configuration explicitly excludes Bash, and the scripts avoid using shell=True or dynamic code execution (e.g., eval or exec).
[PROMPT_INJECTION]: The instructions and examples focus purely on numerical analysis and do not contain patterns typical of prompt injection or behavioral overrides.

linear-solvers