dreaming
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill relies on the 'signet' CLI for all ontology operations, which is consistent with legitimate vendor-provided tools for the Signet-AI ecosystem.\n- [SAFE]: Comprehensive safety constraints are in place, including prohibitions on direct database manipulation and requirements for source-backed evidence for all graph mutations. This minimizes the risk of unauthorized or unverified changes.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to the processing of external transcripts and imported records.\n
- Ingestion points: raw transcripts and source records identified in SKILL.md.\n
- Boundary markers: Absent for untrusted data ingestion.\n
- Capability inventory: Mutating the knowledge graph via the 'signet' CLI.\n
- Sanitization: Relies on auditing, provenance tracking, and the review standard for evidence validation.
Audit Metadata