The Agent Skills Directory

[SAFE]: The skill is designed to execute local modeling scripts (Python or Matlab) to perform perturbations and error analysis. This command execution is constrained to the project's code directory and is necessary for its primary scientific purpose.
[SAFE]: The skill implements a strict "Human Decision Artifact" system (Gate G4.5). It computes data and provides suggestions but explicitly stops and waits for a human to commit a rationale and verdict before the workflow can proceed. This prevents the agent from autonomously validating its own results.
[SAFE]: No evidence of obfuscation, hardcoded credentials, network exfiltration, or persistent mechanisms was found. The skill operates exclusively on internal workspace data and result files.

robustness-checker