What problem does it solve? Everstream Analytics is a read-heavy risk intelligence platform where the real dangers are not posting money or stock but leaking sensitive sub-tier supplier data, silently re-scoring a whole tenant, or acting irreversibly on a probabilistic score. This Skill gives an AI agent the operator judgment to classify every Everstream action by data sensitivity and egress, and to gate consequential ones for human approval. ## Core Features & Use Cases - Action classification by sensitivity and egress: Distinguishes safe in-boundary reads from committing writes (scorecard weight changes, edge curation, alert adjudication) and destructive acts (egress of sub-tier maps, named-supplier scores, or forced-labor findings). - Confidence-aware reasoning: Teaches that risk scores are predictive estimates, sub-tier edges are inferred with confidence levels, and alerts are early signals, so no irreversible action is taken on a single unconfirmed read. - Cross-system gating: Handles pushes of risk flags back into ERP or procurement systems, gating them against the target system's automation rules. - Use Case: A user asks whether to de-source a supplier flagged with a high risk score. The Skill directs the agent to check the score's category, freshness, and confidence, corroborate the signal, and route any supplier block to the ERP under its own approval rules instead of acting on the raw number. ## Quick Start Ask the agent to assess whether a supplier risk flag from Everstream justifies blocking that supplier in SAP, and have it walk through confidence, freshness, and egress checks first.
