databases
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill's setup instructions for PostgreSQL in
SKILL.mdguide the user to execute commands with elevated privileges usingsudoto install database packages and manage system services (sudo apt-get install postgresql,sudo systemctl start postgresql). - [COMMAND_EXECUTION]: The skill facilitates the use of database management command-line tools such as
psql,mongosh, and theatlas cli. Additionally, it instructs the agent on the execution of local Python utility scripts (scripts/db_migrate.py,scripts/db_backup.py,scripts/db_performance_check.py) for administrative tasks like migrations and backups. - [INDIRECT_PROMPT_INJECTION]: As the skill enables the agent to query and process data from external MongoDB and PostgreSQL databases, it presents an attack surface for indirect prompt injection if those databases contain attacker-controlled content.
- Ingestion points: Query results from database read operations and aggregation pipelines (e.g.,
find,SELECT) described in reference files likemongodb-crud.mdandpostgresql-queries.md. - Boundary markers: The provided instructions do not specify boundary markers or include warnings to ignore embedded instructions within data retrieved from databases.
- Capability inventory: The skill grants capabilities to execute database commands, run local Python scripts, and perform file system operations for backup retention.
- Sanitization: There are no documented steps for sanitizing, escaping, or validating external database content before it is analyzed by the agent.
- [EXTERNAL_DOWNLOADS]: The documentation includes instructions for installing official database software and CLI tools from well-known registries using standard package managers (
apt,npm), includingpostgresqlandmongodb-atlas-cli.
Audit Metadata