yuzhaopeng-up

Enterprise WeChat template_card builder and sender. Provides reusable card templates for due diligence reports, stock alerts, risk warnings, and more. Supports text_notice, news_notice, and button_interaction card types with fallback to markdown.

本技能为进出口企业提供全面的贸易融资方案推荐与对比分析，基于企业类型、贸易类型、融资金额、账期等关键参数，智能推荐最适合的融资产品。

Financial AI Skill - 客户经理营销话术实时生成器。输入客户画像和营销目标，AI自动生成电话/微信/拜访话术，支持异议处理预演、方言适配、多风格切换。覆盖零售/对公/理财/私行/信用卡全营销岗位。

网点分析技能是一个基于AI的银行网点竞争力评估与经营规划引擎。该技能综合分析网点的地理位置、竞争格局、客群特征等维度，输出SWOT分析、三年经营预测、重点发展业务方向及投入产出建议。

ALM（Asset-Liability Management，资产负债管理）引擎专注于银行资产负债表的全方位分析与优化。输入银行资产负债数据（资产规模、负债结构、期限分布），输出完整的缺口分析、风险指标（LCR/NSFR/久期缺口）及优化建议。

Financial AI Skill - 产品手册智能对话引擎。基于 BM25 + TF-IDF 双路检索 + RRF 融合的轻量级 RAG，零外部依赖、毫秒级响应、自动出处标注。支持理财/信用卡/贷款等多种产品手册同时入库，语音/文字提问秒回精准答案。Hit@1 实测 100%，平均检索耗时 < 3ms。

Financial AI Skill - 财富管理智能体。提供资产配置、财务健康诊断、退休规划、保险规划、税务优化、教育金规划、房产规划、智能投顾等8大能力。基于规则引擎的轻量级财富管理系统，零API费用。

金融AI技能 - 基金/理财估值核算辅助引擎。输入持仓数据和净值日期，输出估值计算结果与异常预警。验收标准：估值准确率≥99.99%。

Multi-channel health monitoring for multi-agent clusters. Monitors node liveness across all communication channels (P2P relay, message bus, group chat bots), detects heartbeats, tracks version drift and task queue depth, and triggers tiered alerts. Acts as the cluster's "vital signs dashboard".

Decentralized async task handoff for multi-agent clusters using GitHub as a serverless task queue. Agents create/claim/complete Issues as "work tickets", push/pull artifacts via Git branches -- no real-time connectivity required. Ideal for cross-timezone collaboration, CI/CD pipeline handoffs, and any scenario where agents operate on different schedules.

Group chat bot collaboration skill for multi-agent clusters. Enables bots to interact in group chat platforms (Feishu/Lark, DingTalk, Slack, etc.) with human-visible, rich-media collaboration process. Includes the critical "reply tag trap" solution and bot-to-bot negotiation patterns.

Full-async Redis message bus for multi-agent clusters. Provides Pub/Sub broadcast, Stream persistent queue, service discovery, and heartbeat detection. Built on asyncio + aioredis. Ideal for agent clusters that need one-to-many and many-to-many communication with guaranteed message delivery. Industry scenarios: financial analysis workflows, game server orchestration, IoT device coordination, SaaS multi-tenant task dispatch.

End-to-end encrypted P2P communication protocol for multi-agent clusters. Supports relay-based message routing, webhook push (no polling needed), and cross-firewall/NAT traversal. No public IP required on any node. Ideal for sensitive data exchange, private task delegation, and any scenario requiring confidentiality between specific agent pairs.

Natural Language to Query — converts natural language queries into structured query objects. Includes: intent recognition, entity extraction, query construction, confidence scoring, query optimization. Depends on: Info-Extractor, Data-Analyst, Security-Guard. Orchestrated via Phase-Orchestrator. Triggers: natural language query, NL2SQL, text-to-SQL, ask data, smart query.

Visualization Renderer — converts structured data into interactive ECharts charts and dashboards. 4-Phase forced orchestration: data feature analysis -> ECharts config generation -> HTML page rendering -> dashboard multi-chart layout. Depends on diagram-drawing and web-artifacts-builder components, orchestrated via Phase-Orchestrator. Triggers: chart display, visualization, bar chart, line chart, pie chart, dashboard, kanban, data display.

Data Aggregator — performs secondary processing on raw data from a data executor: validation, cleaning, aggregation, year-over-year/month-over-month comparison, statistical enhancement and annotation. 4-Phase forced orchestration via Phase-Orchestrator. Depends on Data-Analyst component. Triggers: aggregation, statistics, grouping, YoY/MoM, TOP ranking, data summarization.

Multi-source evidence chain analysis. Extracts evidence from multiple independent sources, cross-validates, detects conflicts, evaluates confidence, and outputs analysis reports with root cause judgments. Use when: (1) analyzing client complaints against system records to find truth, (2) cross-validating data from 2+ sources (complaints, alerts, SLA agreements, operation logs), (3) detecting contradictions between sources, (4) producing confidence-scored root cause judgments. Triggers: evidence chain, cross-validation, multi-source analysis, conflict detection, confidence assessment, root cause inference, fault diagnosis analysis, complaint verification, alert correlation, evidence chain analysis.

Configurable rule-based scoring engine for multi-dimensional weighted evaluation. Rules are parameterized in YAML configs — change rules without changing the Skill. 4-Phase orchestrated pipeline: Phase1(Info-Extractor) → Phase2(Knowledge-RAG) → Phase3(Data-Analyst) → Phase4(Report-Generator). Use cases: customer opportunity scoring, churn risk assessment, supplier evaluation, partner tiering, and any multi-dimensional weighted scoring needs. Triggers: scoring, rate, opportunity scoring, customer scoring, churn risk, rule engine, multi-dimensional scoring, weighted scoring, rule hit detection. Activated when a user provides a business object (e.g., customer profile) and requests scoring against configurable rules.