ReasoningBank Intelligence
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill implements a system for recording and retrieving task experiences and strategies which creates an indirect prompt injection surface.
- Ingestion points: The recordExperience, learnPattern, and metaLearn methods described in SKILL.md ingest arbitrary task outcomes and contexts from external task executions.
- Boundary markers: No delimiters or instructions to ignore embedded commands are present in the provided documentation to prevent the model from obeying instructions hidden within recorded experiences.
- Capability inventory: The AdaptiveAgent class example demonstrates that strategies retrieved from the ReasoningBank directly influence agent execution via the executeWithStrategy method.
- Sanitization: The documentation does not specify any validation, escaping, or sanitization for stored experience data before it is used for strategy recommendations.
Audit Metadata