database-monitoring

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes administrative SQL commands for managing database processes, such as pg_terminate_backend, pg_cancel_backend, and KILL CONNECTION. These are standard tools for database administrators to troubleshoot stuck or long-running queries and are consistent with the skill's purpose.
  • [EXTERNAL_DOWNLOADS]: References the postgres_exporter utility from the prometheus-community GitHub organization. This is a well-known, trusted source for database monitoring tools.
  • [PROMPT_INJECTION]: The skill involves reading data from system tables that log user-generated query strings (e.g., pg_stat_activity.query, pg_stat_statements.query, and INFORMATION_SCHEMA.PROCESSLIST). This creates a surface for indirect prompt injection.
  • Ingestion points: System views containing query text generated by external database users (SKILL.md).
  • Boundary markers: None present; the SQL queries retrieve raw text from logs without delimiters.
  • Capability inventory: The skill allows for the execution of administrative SQL commands, including process termination and table maintenance.
  • Sanitization: No sanitization or 'ignore instructions' warnings are provided for the analysis of logged query text.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 05:18 PM