The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches materials data from the official Materials Project API (api.materialsproject.org) using the mp-api library. The skill enforces the use of the official endpoint and does not allow redirection of credentials to third-party hosts.
[DATA_EXFILTRATION]: Implements proactive secret management. The scripts/mp_query.py and scripts/_common.py files contain logic to explicitly redact the MP_API_KEY from error messages and logs. Additionally, the scripts/artifact_manifest.py utility refuses to process files with names that suggest they contain credentials or secrets (e.g., 'env', 'secret', 'api-key').
[DYNAMIC_EXECUTION]: Implements a dedicated security control in scripts/_common.py (load_structure_json) to prevent arbitrary Python object instantiation. The script recursively scans structure JSON files to reject 'MSON' metadata keys (@module and @class) before processing, effectively mitigating potential remote code execution vulnerabilities associated with the underlying library's deserialization features.
[COMMAND_EXECUTION]: All bundled CLI scripts use strict file system operations. The scripts/_common.py helper enforces bounds on file sizes, rejects symbolic links, prevents directory traversal via '..' checks, and ensures that output files cannot overwrite input artifacts.

pymatgen