ag2-structured-output
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied strings through the agent interface. While it does not use explicit delimiters in the prompt templates, it implements robust mitigation through Pydantic schema validation which enforces strict adherence to expected data types.
- Ingestion points: agent.ask() in recipe_builder.py.
- Boundary markers: Not explicitly defined in example prompts.
- Capability inventory: Restricted to the scale_ingredient mathematical function.
- Sanitization: Implements robust Pydantic validation and automatic retry logic via the response_schema parameter.
- [SAFE]: The skill utilize vendor-owned libraries (ag2) and interacts with well-known LLM provider APIs for inference.
- [SAFE]: The skill follows security best practices by providing documentation on common pitfalls, such as infinite loops during retries, and demonstrating proper validation patterns.
Audit Metadata