What problem does it solve? Bank customers often leave large balances earning 0% on current accounts or miss products that fit their spending, and support agents rarely surface these opportunities unprompted. This Skill lets the agent proactively spot such optimisations while strictly asking permission first and never acting without explicit confirmation. ## Core Features & Use Cases - Silent opportunity detection: Calls detect_opportunities on product data tools to find idle-cash opportunities with an estimated_annual_gain, plus credit-card candidates when the customer holds no card. - Permission-first recommendations: States a one-line observation, asks "May I make two quick suggestions?", and only then presents two numbered, personalised recommendations grounded in tool figures and cited product conditions. - Human-in-the-loop action: Hands over to the product-ordering skill with a confirm=false preview, verbal confirmation, then confirm=true commit, while applying compliance guardrails on eligibility, segment, and suitability. - Use Case: A customer with €39,000 on a 0% current account asks about a transaction; after answering, the agent flags the idle cash, gets consent, and recommends a savings product with the concrete annual gain and a card cost/benefit based on real spending from summarize_spending. ## Quick Start Ask the agent to check whether there is anything worth optimising on my accounts and, if so, to walk me through two personalised suggestions before doing anything.
