The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it ingest untrusted external data (crystallographic files) and possesses capabilities to write files and perform network operations. While structured scientific data is less prone to typical injection than natural language, the lack of boundary markers during data interpolation into the agent's context qualifies as a low-severity risk.
Ingestion points: The scripts scripts/structure_analyzer.py and scripts/structure_converter.py utilize Structure.from_file() to read user-provided files.
Boundary markers: Absent; the skill does not wrap the contents of processed files in delimiters or provide 'ignore instructions' warnings when presenting analysis to the agent.
Capability inventory: The skill can perform network requests to the Materials Project API (phase_diagram_generator.py) and write arbitrary structure files to the filesystem (structure_converter.py).
Sanitization: Relies on the internal parsing logic of the pymatgen library to validate file structures.

pymatgen