The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill makes extensive use of sudo to perform system-level tasks such as package installation, service management, and modifying restricted configuration files. While these actions are consistent with database administration, they grant the agent significant control over the host system.
[COMMAND_EXECUTION]: The skill executes multiple administrative commands, including systemctl, apt, dnf, psql, and find with destructive flags (-delete) to clear data directories during replication setup. These operations are powerful and require careful execution.
[INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection by processing external or system-generated data.
Ingestion points: The agent ingests data from database monitoring queries (pg_stat_activity, pg_stat_user_tables) and local initialization scripts (init.sql) used in Docker environments (SKILL.md).
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat this data as untrusted or separate from its core logic.
Capability inventory: The skill possesses the capability to execute shell commands, modify database permissions, and perform network-related database operations across various system files.
Sanitization: No validation or sanitization of the database output or SQL file content is performed before the agent processes it.

postgresql-devsec