What problem does it solve? Real-time visibility platforms like project44 are mostly read-only, so the real risks are not stock or money postings but sensitive shipment, location, and customer data leaving the trust boundary, and irreversible decisions made on unconfirmed predicted ETAs or self-reported carrier milestones. This Skill gives an AI agent the operator judgment to classify every project44 action by data sensitivity and egress, and to gate consequential writes before they bind physical resources or trigger downstream automation. ## Core Features & Use Cases - Egress-aware action classification: Categorizes operations into read, reversible write, committing write, and destructive/high-sensitivity tiers, with explicit gate ladders and a default-to-block rule for unapproved destinations. - ETA and milestone skepticism: Distinguishes planned, carrier-provided, and predicted ETAs; defines a corroboration standard requiring a second independent source before any irreversible action. - Mode-specific depth: Reference files cover truckload, LTL, ocean (LFD, demurrage, discharge vs available), air, parcel, and rail milestones, EDI 214 codes, and freshness expectations per mode. - Use Case: A user asks for ETAs on all of a customer's containers on a lane. The Skill classifies this as lane/network-composition data, checks the asker's party entitlement, prefers an aggregate answer over named-shipment detail, and blocks export to unapproved destinations. ## Quick Start Ask the agent to check the predicted ETA and exception status of a shipment tracked in project44 and advise whether it is safe to notify the customer.
