nmt-chat
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user context such as half-formed ideas, chat threads, or documents, and it performs live web enrichment. This ingestion of external data creates a standard surface for indirect prompt injection, where instructions hidden in the processed materials could attempt to influence the agent's behavior. The skill includes a behavioral mitigation by instructing the agent to 'Show the seams' and clearly label canon-grounded information versus external enrichment.
- [EXTERNAL_DOWNLOADS]: The skill uses network operations to fetch verification data from academic journals (Nature, Cell), business research platforms (Gartner, Forrester), and well-known technology services (Stripe, Slack, Salesforce). While these sources are reputable and aligned with the skill's professional purpose, arbitrary web fetching is a risk factor for data ingestion.
Audit Metadata