The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data returned from database queries, which is an intrinsic part of its functionality. This creates an attack surface where malicious instructions stored within a database could potentially influence the agent when read. * Ingestion points: Data from SQL query results (e.g., 'projects/postgres/src/db.rs', 'projects/postgres/src/tools.rs'). * Boundary markers: Results are returned as structured JSON or rendered tables, providing separation between data and agent instructions. * Capability inventory: The skill can execute SQL, modify session settings, and manage configuration (e.g., 'projects/postgres/src/main.rs'). * Sanitization: Implements automated redaction of credentials in URLs and error messages to prevent accidental exposure.
[EXTERNAL_DOWNLOADS]: The documentation search feature in 'projects/postgres/src/docs.rs' initiates network requests to 'https://www.postgresql.org/search/'. This is a well-known service used legitimately for retrieving official database documentation.
[COMMAND_EXECUTION]: The skill executes a bundled local binary for database operations. The complete source code for this binary is included in the skill package under 'projects/postgres/', allowing for verification and local builds using provided maintenance scripts. The launcher script 'scripts/postgres' securely selects the platform-specific binary based on system architecture.

postgres