bdi-mental-states

Pass

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill contains architectural documentation, RDF/Turtle examples, and SPARQL queries intended for modeling agent mental states. No malicious commands, unauthorized network access, or obfuscation techniques were detected.
  • [PROMPT_INJECTION]: The skill defines a pattern for processing external data which represents a surface for indirect prompt injection.
  • Ingestion points: External RDF context and world state notifications defined in SKILL.md.
  • Boundary markers: The Python pseudocode example uses simple string interpolation without explicit boundary markers for the external ontology context.
  • Capability inventory: The framework involves generating responses via an LLM and extracting RDF triples for validation.
  • Sanitization: The skill incorporates consistency validation logic to check generated triples against the source ontology before completion.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 10, 2026, 12:28 AM
Security Audit — agent-trust-hub — bdi-mental-states