plan-rls-audit
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (database schemas and Row-Level Security policies) to generate audit reports, creating a surface for indirect prompt injection.\n
- Ingestion points: The skill reads database metadata, policy definitions, and project files (migrations, client bundles) to perform the audit.\n
- Boundary markers: The instructions lack formal delimiters for untrusted data, though they advise the agent to "read each policy's truth table, don't trust its name."\n
- Capability inventory: The skill is strictly limited to planning and reporting. It contains explicit guardrails ("Plan only. No SQL runs") prohibiting state-changing operations like
ALTER TABLEorCREATE POLICY.\n - Sanitization: There is no mention of escaping or validating the content of RLS policies or database schemas before including them in the final markdown report.
Audit Metadata