bdi-mental-states

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security vulnerabilities were detected in the skill instructions or reference files. The content is focused on providing a semantic framework for cognitive modeling and agentic reasoning.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting external RDF triples into the agent's mental model via the Triples-to-Beliefs-to-Triples (T2B2T) paradigm. While this architectural pattern creates a surface for indirect prompt injection via untrusted external data, the skill includes proactive mitigation strategies such as Logic Augmented Generation (LAG) and strict ontology-based validation of generated triples to ensure the agent's mental state remains consistent and secure.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 09:38 AM
Security Audit — agent-trust-hub — bdi-mental-states