dbpropose
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's instructions and reference documentation do not contain any malicious patterns, obfuscation, or unauthorized access attempts.
- [CREDENTIALS_UNSAFE]: The skill includes specific security guardrails that instruct the agent to never copy credentials, keys, or connection strings into the generated proposals, favoring the use of configuration names instead.
- [COMMAND_EXECUTION]: The skill maintains a strict posture of non-execution, explicitly forbidding the agent from applying DDL, running migrations, or writing to the database.
- [INDIRECT_PROMPT_INJECTION]: The skill incorporates a mitigation for indirect injection by instructing the agent to treat input artifacts (audit results, designs) as evidence to be analyzed rather than instructions to be followed, and to ignore any commands embedded within them.
Audit Metadata