The Agent Skills Directory

[Data Exposure & Exfiltration] (SAFE): The script phase_diagram_generator.py correctly retrieves the Materials Project API key from an environment variable (MP_API_KEY), avoiding hardcoded credentials. Network activity is limited to fetching data from the legitimate Materials Project API.
[Indirect Prompt Injection] (SAFE): While the tools ingest untrusted structure files (CIF, POSCAR, etc.) in structure_analyzer.py and structure_converter.py, the data is parsed into strict materials science objects. There is no evidence that the agent would execute instructions embedded in these data files.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill relies on well-known scientific libraries (pymatgen, mp-api). No patterns of remote script execution or arbitrary command execution were found.
[Dynamic Execution] (SAFE): The skill promotes secure serialization practices using as_dict() and from_dict() for JSON/YAML persistence, explicitly avoiding the security risks associated with Python's pickle module.

pymatgen