dbeaver
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill contains configuration examples that reference sensitive filesystem locations used for authentication and encryption.
- Evidence: The 'Configuring JDBC Connection' section in
SKILL.mdexplicitly references~/.ssh/bastion_key.pemfor SSH tunneling. - Evidence: The 'Best Practices' section mentions
dbeaver-data-sources.xml, a configuration file known to store database credentials. - [INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to assist with database administration and query execution, which involves processing external, potentially untrusted data that could contain malicious instructions.
- Ingestion points: Database query result grids and exported datasets (JSON, CSV, SQL) as described in the 'Exporting Datasets' section of
SKILL.md. - Boundary markers: None present; the instructions do not specify delimiters or provide warnings to the agent to ignore instructions embedded within database records.
- Capability inventory: SQL execution, CLI script execution, and data migration/transfer between heterogeneous database engines.
- Sanitization: No sanitization, filtering, or escaping logic is defined for handling content retrieved from database tables.
Audit Metadata