plsql-write

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFE
Full Analysis
  • [DYNAMIC_EXECUTION]: The instructions explicitly address the risks of dynamic SQL execution. They warn against SQL injection through string concatenation and mandate the use of bind variables via the USING clause as a required security pattern.\n- [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by reading existing codebase source and metadata using the pythia tool. However, the risk is mitigated as the skill emphasizes human review for style changes and provides strict security patterns for the generated output, such as validating identifiers against system views (ALL_TABLES).\n
  • Ingestion points: Reads codebase source and metadata through pythia (SKILL.md).\n
  • Boundary markers: None explicitly mentioned for isolating external tool output.\n
  • Capability inventory: File system write access for PL/SQL source files.\n
  • Sanitization: Mandates bind variables and identifier validation for generated code.\n- [SAFE]: The skill enforces a principle of least privilege for tool usage by directing the agent away from powerful, general-purpose SQL execution tools like sqlplus in favor of a specific application utility (plsql-apply).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 09:08 AM
Security Audit — agent-trust-hub — plsql-write