The Agent Skills Directory

EXTERNAL_DOWNLOADS (SAFE): The skill mentions the pscale CLI and the pg_activity Python package. These are standard, legitimate tools for database administration and monitoring.\n- DATA_EXPOSURE (SAFE): All connection strings and credential examples use placeholders (e.g., <user>, <password>, <token>), ensuring no sensitive data is exposed.\n- INDIRECT_PROMPT_INJECTION (LOW): The agent is instructed to ingest schema and performance data from PlanetScale APIs. Safety instructions are included throughout the documentation to require human confirmation before executing any destructive commands, such as dropping indexes or partitions.

postgres