database-testing

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION] (HIGH): The skill utilizes functions like db.raw() for arbitrary SQL execution and migrate()/rollback() for schema modification. These capabilities can be hijacked to perform destructive database operations if input parameters are not strictly validated.\n- [PROMPT_INJECTION] (HIGH): The skill is vulnerable to Indirect Prompt Injection (Category 8) due to its high-privilege capabilities combined with untrusted data ingestion.\n
  • Ingestion points: Parameters in Task() calls (e.g., schema, migration) are potential injection vectors from external sources.\n
  • Boundary markers: No delimiters or isolation instructions are present to prevent embedded commands in parameters.\n
  • Capability inventory: Includes data deletion (db.users.delete), schema rollback, and raw SQL execution.\n
  • Sanitization: No evidence of input validation, escaping, or allow-listing exists in the provided implementation.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 08:43 AM