semantic-relations

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely instructional, focusing on semantic relation types such as IS-A, PART-OF, and synonymy. No scripts, binaries, or automated tool invocations (besides standard Read/Grep) are present.
  • [EXTERNAL_DOWNLOADS]: The skill references several external documentation and educational resources, including Princeton WordNet, W3C standards (SKOS, OWL), and FrameNet. These are well-known technology and academic sources used for grounding knowledge and do not involve remote code execution.
  • [DATA_EXFILTRATION]: No network operations or sensitive file access patterns were detected.
  • [PROMPT_INJECTION]: The instructional content does not contain attempts to override agent behavior or bypass safety guardrails.
  • [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process user-supplied concept names and descriptions for relational analysis, it lacks dangerous capabilities (such as shell execution or external network writes) that would create an exploitable attack surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 04:35 AM
Security Audit — agent-trust-hub — semantic-relations