What problem does it solve? Working with Interos risk intelligence is dangerous in non-obvious ways: the graph is AI-modeled rather than attested, scores can be read on the wrong scale, and the real risk is sensitive sub-tier and sanctions data leaving the trust boundary. This Skill gives an AI agent the operator judgment to classify every Interos action by data sensitivity and egress, and to stop consequential actions for human approval. ## Core Features & Use Cases - Action classification by sensitivity and egress: Gates reads, watchlist writes, tenant-wide scorecard changes, graph curation, and cross-system pushes on a four-tier egress ladder instead of treating everything as a safe read. - Graph and score skepticism: Teaches verification of entity resolution, edge confidence, score direction (resilience vs risk), factor breakdowns, and feed freshness before acting on an i-Score or propagated alert. - Blind-spot and destructive-by-proxy protection: Prevents muting alerts or dropping watchlist entities to quiet noise, and hard-gates pushing risk flags into ERP/procurement systems that may auto-block suppliers. - Use Case: A user asks to email a shared-node concentration map to an outside consultant; the Skill classifies the payload as high-sensitivity egress, requires a named owner and approved destination, and defaults to blocking when the destination is unknown. ## Quick Start Ask the agent to assess a supplier's i-Score and concentration exposure in Interos and recommend whether any action should be gated for human approval.
