The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes SQL queries through a Python wrapper. Safety is maintained by enforcing readonly=True at the PostgreSQL session level and performing client-side checks to allow only SELECT, SHOW, EXPLAIN, and WITH statements.
[DATA_EXPOSURE]: The skill manages database credentials (host, user, password) via a local JSON configuration file. It includes logic to check and warn about insecure file permissions (Unix chmod 600) and sanitizes authentication-related errors to prevent credential leakage in logs.
[INDIRECT_PROMPT_INJECTION]: As the skill reads and displays data from external databases, it is subject to indirect prompt injection if the database records contain malicious instructions. The skill mitigates large-scale data ingestion issues by enforcing a MAX_ROWS limit (10,000) and truncating long column values.

postgres