The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and analyze potentially untrusted data from database query logs and system statistics, which could contain risky patterns intended to influence the agent's behavior.
Ingestion points: Data enters the context through database tools querying system views like pg_stat_activity and query_store.qs_view, which contain query strings executed by external users.
Boundary markers: The skill does not explicitly use delimiters or instructions to ignore embedded commands within the text retrieved from the database.
Capability inventory: The agent has the ability to modify the database (postgres_mcp_modify) and execute shell commands via the Azure CLI, which are potential targets for an injection attempt.
Sanitization: No specific sanitization or pre-processing of database results is described in the provided instructions.
Context: While this ingestion is essential for the skill's diagnostic and tuning purposes, the potential for impact is mitigated by strict policies requiring explicit user verification before the agent executes any destructive or administrative operations.
External Resource Reference: The skill refers to a well-known external service for security-related configurations.
Details: The documentation in azure-postgresql-networking-ssl.md suggests downloading a root certificate from the DigiCert certificate authority's official domain. This is a secure, industry-standard practice for establishing encrypted and verified database connections.

postgresql-best-practices