The Agent Skills Directory

Managed Identity Authentication: The skill demonstrates the use of DefaultAzureCredential, which is a security best practice for Azure services as it avoids the need for hardcoded secrets and supports various authentication environments.
Injection Mitigation: The reference guides explicitly warn against string interpolation in queries and provide detailed examples of parameterized queries. This follows the principle of least privilege and protects against query injection.
Official Package Usage: The installation instructions utilize official, versioned packages (azure-cosmos, azure-identity) from Microsoft, ensuring the use of trusted code for interacting with Azure services.
Data Handling Considerations: The skill processes data from external database sources. While this presents a potential indirect injection surface, the implementation mitigates this through safe practices:
Ingestion points: Data enters via container.read_item in SKILL.md and container.query_items in references/query-patterns.md.
Boundary markers: The skill consistently uses parameterized queries to separate data from instructions.
Capability inventory: The skill is scoped to database operations using the standard azure-cosmos library.
Sanitization: The instructions emphasize parameterization and the use of SDK-provided methods for safe data retrieval and handling.

azure-cosmos-py