database-administrator

Warn

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill contains numerous Bash scripts designed for automated database maintenance, including 'pg_full_backup.sh', 'postgresql_upgrade.sh', and 'zero_downtime_migration.sh'. These scripts perform low-level operations such as database dumps, service restarts, and file system management using the Bash tool.
  • [PRIVILEGE_ESCALATION]: The provided scripts utilize 'sudo -u postgres' and package managers like 'apt-get' to perform administrative tasks, which involve executing commands with elevated permissions.
  • [DATA_EXFILTRATION]: Scripts such as 'backup_monitor.sh' and 'pg_full_backup.sh' use 'curl' to transmit operational status and notifications to external Slack webhooks. Additionally, 'aws s3 cp' is used to transfer database backups to remote storage buckets.
  • [EXTERNAL_DOWNLOADS]: The skill uses standard system tools like 'apt-get' to fetch and install database software packages from official repositories during upgrade procedures.
  • [PROMPT_INJECTION]: The instructions use strong imperative language ('CRITICAL', 'MANDATORY', 'ALWAYS') to override standard agent behavior, forcing the agent to prioritize and adopt instructions found in external files located in the 'steering/' directory.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: The skill is instructed to read context from 'steering/structure.md', 'steering/tech.md', and 'steering/product.md' before starting tasks. It also analyzes database slow logs and query execution plans.
  • Boundary markers: There are no explicit delimiters or warnings to ignore embedded instructions within the steering files.
  • Capability inventory: The skill has access to 'Bash', 'Read', 'Write', and 'Edit' tools, allowing it to execute scripts that perform network requests and filesystem modifications.
  • Sanitization: There is no evidence of sanitization or content validation for the data ingested from the steering files or database logs.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 1, 2026, 02:15 AM