Skills 导航

聚焦生产级数据工程，覆盖 ETL/ELT、批处理与流式管道、数据建模、Airflow/dbt/Spark 优化和数据质量治理，适合设计数据架构、搭建现代数据栈与排查性能问题。

✎ 复杂数据管道、ETL/ELT 和治理难题交给它，凭 Spark、Airflow、dbt 等现代数据栈经验，能更稳地搭起可扩展的数据基础设施。

覆盖Prompt优化、Few-shot设计、结构化输出、RAG评测与Agent工作流编排，适合分析token成本、评估LLM输出质量，并搭建可落地的AI智能体系统。

✎ 把提示优化、LLM评测到RAG与智能体设计串成一套方法，适合想系统提升AI开发效率的人。

覆盖模型部署、MLOps 流水线、特征库、漂移监控与 RAG/LLM 集成，适合把训练好的模型稳定上线，并持续优化性能、成本与迭代效率。

✎ 想把模型真正跑进生产环境，ML工程师把部署、MLOps、RAG到漂移监控串成完整闭环，连成本优化都替你想到。

Write high-performing, persuasive, and authentic LinkedIn posts across any professional niche. Uses research-backed hooks, proven post structures, and LinkedIn-specific formatting to maximize reach, saves, comments, and profile actions. Use when creating LinkedIn posts, carousels-by-text, thought leadership, founder updates, case studies, hiring posts, launch posts, or converting X/Twitter ideas into LinkedIn content.

Convert long-form content into platform-native assets for LinkedIn, X, newsletter, and email while preserving message consistency and brand voice.

Multi-channel notification delivery with customizable templates, email and SMS sending, and full delivery logging — built for AI agents.

Event tracking, funnel analysis, cohort retention, multi-touch attribution, revenue metrics, and real-time analytics — built for AI agents.

Programmatic backtesting framework for trading strategies. Runs backtests with historical price data (yfinance or CSV), supports momentum/mean-reversion/factor/signal-based strategies, walk-forward optimization, out-of-sample testing, transaction cost modeling, regime-aware splits, and full performance metrics (Sharpe, Sortino, Calmar, max drawdown, CAGR, win rate, profit factor). Distinct from einstein-research-backtest (which provides methodology guidance). Use when a user wants to actually run a backtest, test a specific strategy on historical data, or generate performance metrics.

Configure and use Volcano Engine (Volcengine) models including Doubao, GLM, Kimi, and DeepSeek. Use when: (1) Setting up Volcengine API access in OpenClaw, (2) Choosing between general vs coding endpoints, (3) Configuring model aliases for easy access, (4) Troubleshooting authentication or connection issues with Volcengine providers.

聚焦高级 Prompt 设计与优化，覆盖 few-shot、chain-of-thought、模板体系和 system prompt，适合为复杂推理任务构建可复用、可评估、可上线的生产级提示方案。