bdi-mental-states
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a 'Triples-to-Beliefs' pattern that ingests external RDF context to update the agent's internal mental states. \n
- Ingestion points: External RDF triples and context are parsed to form beliefs (SKILL.md). \n
- Boundary markers: The skill includes an explicit instruction to 'keep untrusted RDF as data' and notes that framework mappings are illustrative. \n
- Capability inventory: The skill is restricted to ontological modeling and does not define any capabilities for file writes, network requests, or subprocess execution. \n
- Sanitization: The provided pseudocode includes a 'validate_triples' step to ensure consistency against a verified ontology graph. \n- [SAFE]: The skill provides guidelines for BDI modeling and cognitive architecture. It contains no executable scripts, hardcoded credentials, or instructions that bypass safety filters. All Python and Prolog snippets are provided as pseudocode or examples.
Audit Metadata