The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate the ingestion of external data from a PostgreSQL database into the agent's execution context through query results.
Ingestion points: Data retrieved via client.query or pool.query (e.g., in SKILL.md under core workflows).
Boundary markers: Not explicitly defined for the LLM context, though standard database result objects are used.
Capability inventory: The skill facilitates network operations (database connections) and file-system interactions (reading environment variables).
Sanitization: The skill actively promotes security by mandating parameterized queries ($1, $2) to sanitize inputs directed at the database, significantly reducing the risk of SQL injection.
[EXTERNAL_DOWNLOADS]: The skill references standard, well-known libraries from the official NPM registry.
Evidence: npm install pg @azure/identity in SKILL.md. These are widely recognized, industry-standard packages for database connectivity and Azure authentication.
[CREDENTIALS_UNSAFE]: The skill provides instructions for managing secrets through environment variables.
Evidence: References to AZURE_POSTGRESQL_PASSWORD and .env files. This follows standard secure development practices by using placeholders rather than hardcoded credentials.

azure-postgres-ts