problem-details-rfc9457
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a structured framework for the agent to generate Java exception handlers.
- Ingestion points: The agent processes user requirements and domain models to generate Spring Boot code.
- Boundary markers: The skill provides predefined Java templates (templates/*.java) that serve as structural constraints for the output.
- Capability inventory: The agent modifies and creates Java source files within the project workspace.
- Sanitization: The instructions and templates (SKILL.md, good-exception-handler.java) explicitly mandate the sanitization of error details, ensuring internal exception messages are logged privately and only generic responses are returned to API consumers.
- [SAFE]: Comprehensive analysis of the instructions, templates, and examples found no evidence of malicious behavior, obfuscation, or unauthorized access. The skill includes negative examples to specifically train the agent to avoid common security pitfalls like information leakage via stack traces.
Audit Metadata