The Agent Skills Directory

[Data Flow and Connectivity]: The skill connects to Azure AI Foundry endpoints for model inference. These operations are performed over secure HTTPS connections using Microsoft's standard Entra ID authentication (via DefaultAzureCredential), ensuring that no sensitive credentials are hardcoded and data is transmitted securely to legitimate vendor services.
[Dependency Management]: The skill utilizes well-known and reputable scientific and web development packages from official registries (PyPI and NPM). No suspicious or unverifiable dependencies were identified in the project's configuration files.
[Data Handling]: While the skill processes external crystal structure files (CIF/XYZ), it uses established libraries like pymatgen for parsing. The data ingestion surface is used strictly for its primary scientific purpose, and the specialized nature of the diffusion models involved (MatterGen) prevents traditional prompt injection vulnerabilities associated with general-purpose language models.
[Container Security]: The provided Dockerfiles and startup scripts follow secure defaults, such as running with non-root users where appropriate and correctly managing permissions for essential services like Nginx within the isolated container environment.
[Information Integrity]: All documentation and metadata provided are consistent with official Microsoft Research publications and repositories, confirming the skill's legitimate research and development purpose.

msresearch-mattergen