RoacherM

获取A股数据(baostock)并缓存到本地CSV文件，避免MCP返回大量数据占用上下文。触发场景：(1)获取超过100条的K线数据 (2)需要多次查询同一股票数据 (3)需要用grep/awk分析数据 (4)用户提到"保存数据"或"缓存数据"

执行威科夫市场结构分析并生成交互式图表。触发场景：(1)分析股票吸筹/派发形态 (2)识别威科夫阶段(Phase A-E) (3)标注SC/AR/Spring/SOS等事件 (4)用户提到"威科夫分析"或"Wyckoff"

Git-native OKR tracking with AI intelligence. Use when: (1) initializing OKR tracking ('set up OKR', 'init okr', '帮我建OKR'), (2) recording progress in natural language ('retention improved to 45%', 'data platform shipped', 'hiring blocked', '精度提升到90%', '被阻塞了', '完成了'), (3) checking OKR status ('how are we doing', 'show progress', 'dashboard', 'OKR进度怎样', '看板'), (4) revising spec ('adjust target', 'cancel KR', 'freeze KR', '调整目标', '取消KR'), (5) closing a quarter ('close quarter', 'wrap up Q1', '季度关闭'), (6) visualizing progress ('metro map', '可视化', '地铁图'). Triggers: /okr, any KR progress update, metric change, blocker report, okr init/status/revise/close, metro map. Even if user doesn't say 'OKR' explicitly — if they mention KR progress, quarterly goals, or metric updates related to their OKR setup, use this skill.

First-principles architecture thinking using GPT-5.x via OpenRouter. Two modes — "design" generates a system scheme from scratch, "review" critiques an existing design. Use when the user says "deep think", "deep design", "deep review", "/deep-think", "help me design", "review my architecture", or has a system idea or existing design that needs structural thinking. Evaluates through 5 layers (Why → What → How → Quality → Proof). NOT for code style or implementation details.