菜单
Use when the customer asks about their balance, transactions, or personal details. Produces a grounded, confidential answer from the customer data MCP tools.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Operations runbook for the agentic-banking-ecosystem repo. USE THIS SKILL when the user asks to deploy, build, provision, generate data, index, ingest, register, or clean up any part of the project — including infrastructure (azd), container images, Container Apps, the customer/product MCP servers, AI Search indexes, Foundry toolboxes, and the banking agents. Covers the full lifecycle: provision → generate data → build → deploy MCP → register toolboxes → index → ingest → deploy agents → clean up.
Use when an agent asks what data is required for a customer, account or transaction to be compliant — an onboarding checklist, an eligibility gap analysis, or "what do we still need to collect". Maps each compliance rule onto the concrete banking data model and returns an extensive, field-level requirements breakdown.
Use before asking Bank North's Compliance agent (the ask_compliance tool) any regulatory, KYC/AML, sanctions, fraud or product-eligibility question. Gathers the signed-in customer's real data first, then composes a precise, self-contained compliance question so the answer is specific to this customer and product rather than generic.
Use when the customer wants to order a product (e.g. open a savings account or apply for a credit card) or change their contact details. Enforces the human-in-the-loop confirmation and compliance guardrails.
Use when the agent should proactively spot an optimisation for the signed-in customer (e.g. a large uninvested balance on the current account) and offer help. Always asks permission first, then gives two concrete, personalised recommendations, and only acts human-in-the-loop.
Use when an employee asks which product to recommend to a customer, or to compare products for a specific customer need. Produces a grounded, segment-aware recommendation.
| name | account-enquiry |
| description | Use when the customer asks about their balance, transactions, or personal details. Produces a grounded, confidential answer from the customer data MCP tools. |
Answer a customer's questions about their own accounts, balances and transactions, strictly for the signed-in customer.
get_balance / get_account for a single holding balance.get_net_worth for a balance overview across ALL holdings (total + a
breakdown by product type). Prefer this for "all my accounts" / "total"
questions instead of summing manually.summarize_spending for any spending question — it returns the total, a
breakdown by category, the top merchants and the single largest
transaction over the period. Pass date_from / date_to for a window and
category to focus on one spending category.list_accounts / get_customer for the full holding list and profile.list_transactions with date_from / date_to when the customer wants
the raw statement lines rather than a summary.update_overview with the customer profile and
the current accounts before replying.Lead with the direct answer (e.g. the balance), then a compact table of transactions (date, description, category, amount, balance after) when relevant.