skills

Karpathy's LLM Wiki: build/query interlinked markdown KB with Zettelkasten note discipline. Triggers: 知識庫, 建 wiki, wiki 筆記, 研究筆記, ingest sources, knowledge base.

Use when building, maintaining, or auditing a Zettelkasten (slip-box) note system, or when applying atomic-note discipline to an existing markdown knowledge base — atomicity checks, card-splitting decisions, linking rules, note-type workflow. Trigger on 卡片盒筆記法、卡片盒、Zettelkasten、原子筆記、永久筆記、文獻筆記、結構筆記、拆卡、筆記連結、smart notes、atomic notes、permanent notes、slip-box, or when another skill (e.g. llm-wiki) needs per-change note-discipline checks.

Use whenever the agent thinks at all — writing copy or reports, proposals, planning products or features, coding, debugging, or any other task — running ALL curated thinking frameworks in sequence (currently 邏輯推導 logical reasoning：MECE, 空雨傘, 魚骨圖; 創意發想 creative ideation：六頂思考帽, 腳本圖, 心智圖; 市場分析 market analysis：3C 分析, SWOT 分析, 4P與4C; 進度管理 progress management：惠特默模型, KPI 樹狀圖, PDCA; 權衡得失 weighing trade-offs：決策矩陣, PMI 表, 力場分析; 預測未來 predicting the future：S 型曲線, 鴻溝理論, 長尾模型; 辯論思考 debate thinking：需根解損, 判定立駁, 論點以主張-資料-推論三要素構建 Claim-Data-Warrant、以質詢拆解三問檢驗（事實為真？邏輯嚴密？有無其他可能？）; 思維模型 mental models：第一性原理, 反向思考法, 二階思考, 能力圈, 奧坎剃刀) to lock direction and key points before acting, screening all reasoning with a ~160-item 謬誤檢核清單 fallacy checklist. Theme grouping is navigation only — work through every theme, not just the ones that look relevant. Trigger on ultrathink, 深度思考, 想清楚再做, 思考框架, 抓重點, 言之有物, 沒抓到重點, any framework or tool name above, or whenever a task involves thinking, planning, writing, building, or debugging.

Use by default for any user-facing writing, including answers to users, emails, documents, formal correspondence, reports, proposals, landing-page copy, social posts, essays, commentary, explanations, summaries, product copy, internal updates, and other prose intended to be read by people. Apply it whenever the output contains human-readable text, unless another skill explicitly owns the writing style or the output is purely code, structured data, or a command. Supports Chinese and English.

Software engineering guidelines for any software change. Always load this skill before requirement clarification, architecture/design, implementation, refactoring, code review, testing, or shipping. Covers simplicity, surgical changes, TDD, and verifiable success criteria. Triggers on: 任何軟體規劃, 需求釐清, 架構設計, 寫 code, 改 code, 做功能, 修 bug, refactor, 開發, coding, development, 實作, 實現, 寫程式, 改程式, 加功能, 修問題, code review, 測試, 重構

Universal database engineering rules. Migration discipline, audit logging, soft delete, performance, data integrity. Triggers on: 任何資料庫相關工作, 建表, 改 schema, 寫 migration, DB design, database development, 設計資料庫, 改 DB 結構, 加欄位, 加索引, 改 table, SQL, 寫 trigger, 設計 API 的 DB 層, 查詢很慢, N+1, 效能優化, 資料完整性, CASCADE, 外鍵, 新增 table, 刪除資料, backup, 環境分離, dev prod 分開

Use when analyzing a group, business-unit, product-line, or brand portfolio with the BCG growth-share matrix, especially when the task involves relative market share, market growth, quadrant classification, capital allocation, portfolio prioritization, or deciding whether to invest, maintain, harvest, reposition, or exit.

Use when designing, evaluating, or refining a business model with nine elements, especially for new products, pivots, monetization, and go-to-market decisions. Starts with Epicenter Selection (震央選擇) — choosing one of four innovation starting points (資源／產品／顧客／財務導向) to anchor design before filling the nine boxes.

Use when evaluating high-stakes decisions with bias-aware quality control, including proposal review, decision meeting facilitation, and personal coaching. Trigger on requests such as 重大決策、偏誤檢查、提案審查、決策會議引導、決策教練、go/no-go 決策。

Use when conducting external environment analysis for strategic planning, market entry, competitive assessment, or feeding into SWOT. Trigger on requests like PESTEL分析、PEST分析、總體環境分析、外部環境掃描、宏觀環境分析、macro environment、political economic social technological environmental legal、總經分析、產業環境、市場進入分析、or when the user provides a company/industry/product and needs external environment mapped before strategy work. Also trigger when the swot-analysis skill reaches 階段一 and needs PESTEL data as input. Never generate factors without evidence-based sourcing and scoring — every factor must be rated on importance and impact before being carried forward.

對任何契約、條款或法律文件做「對使用者不利條款」掃描,白話翻譯每一條陷阱。適用於各種合約審閱情境:一般商業合約、公寓租約、健身房/補習班/訂閱服務會員合約、工作 Offer/聘僱契約、App 或網站服務條款 ToS、隱私權政策、保險合約、貸款合約、保固條款等。觸發關鍵詞:幫我看合約、審合約、合約紅旗、紅旗條款、檢查合約、租約、會員合約、退會條款、聘僱契約、競業條款、服務條款、ToS、隱私政策、自動續約、隱藏費用、違約金、責任免除、不平等條約、契約陷阱、貸款合約、保險合約。也在以下情況主動觸發:使用者上傳/貼上看起來像契約或條款的文字並問「這份有沒有問題」「我該簽嗎」「有什麼要注意的」「對方有沒有偷塞東西」,或想知道某條款「正常嗎」。預設輸出繁體中文,英文合約原文也能直接讀取分析。

Use when designing, executing, or facilitating a complete SWOT strategic analysis workflow — especially when the task involves environment scanning (PEST/industry analysis), competitive positioning, S/W/O/T identification and validation, strategy formulation via SO/ST/WO/WT collision, and strategy selection. Trigger on requests like SWOT分析、優劣勢分析、機會威脅分析、策略定調、外部環境分析、競爭策略、TOWS矩陣、SWOT到策略、PEST分析、or any combination of competitive analysis and strategy direction. Also trigger when the user uploads a case, company description, or product brief and asks for strategic analysis or positioning.

Use when the final customer journey map, CJM, or touchpoint-based journey table must be produced from a completed persona, product or service, or a `handoff_to_customer_journey_mapper` block, especially when the user asks for 顧客旅程地圖, CJM, final journey table, or wants prior persona/framing work turned into the finished map. Also trigger when the user asks for 情緒分數, 情緒曲線, 加減分原因, sparkline, or wants emotion scoring added to an existing CJM.

Use when building a persona, customer profile, or journey-prep brief from a product, audience clue, or usage context, especially for persona、人物誌、顧客輪廓、使用者輪廓、目標客群、5W1H、接觸點盤點、顧客旅程前置分析、或需要簡報化的顧客研究摘要。

Use when research, analysis, evaluation, diagnosis, discovery, insight, or study-design tasks require deciding between qualitative, quantitative, or mixed-methods approaches, especially when the user does not specify a method or the work needs both measurement and explanation.

Run a product-analysis focused revealed-preference logistic conjoint workflow (product specs + reviews/transactions, stacked binary choice, logistic estimation) — from attribute design through business insights (importance, willingness-to-pay, optimal product configuration, pricing and cost-benefit). Use this skill when the goal is to analyze products and infer multi-attribute preferences from observed market behavior, not a generic survey-only conjoint setup.

Use when customer reviews, support tickets, app store feedback, or other review-like text must be converted into STP analysis through a review scoring workflow upstream and statistical scripts downstream.

Use this skill when the user has product review files and wants to: (1) score each review on a set of attributes using a salience scale (0–7) and export a review × attribute matrix as .xlsx; (2) run PCA to reduce attributes into latent dimensions; (3) cluster reviews into customer segments with K-means. Trigger on requests like 「逐則評論評分」、「顯著度矩陣」、「salience scoring」、「主成分分析」、 「PCA」、「顧客分群」、「K-means」、「評論×屬性 Excel」, or any combination thereof. Also trigger when this step follows an attribute-discovery workflow and the user wants scored data, dimensionality reduction, or segmentation. Use this skill even if the user only mentions part of the pipeline.

Use this skill whenever the user has one or more product review files and wants to (1) extract product attributes from the reviews, (2) score each product on those attributes, and (3) export the results into a Word document (.docx). Trigger on requests like 「從評論歸納屬性」、「幫我評分每個產品」、「做成 Word」、「產品屬性分析」、 「評論轉評分表」, or any combination of "reviews → attributes → scores → Word/docx". Also trigger when the user uploads review files (CSV, JSON, TXT, or similar) and asks for comparative analysis, attribute extraction, or scoring — even without mentioning Maslow or Word explicitly.

Use when coding cross-source evidence with Product Positioning Theory to classify attributes, functions, benefits, and usage/service context, and when optionally exporting theory_annotations compatible with review-mining-stp.

Use when coding cross-source evidence with Purchase Motivation Theory to classify functional, security, and relational drivers, and when optionally exporting theory_annotations compatible with review-mining-stp.

Use when coding cross-source evidence with Word-of-Mouth Motivation Theory to classify altruistic, social identity, self-enhancement, and emotional expression motives, and when optionally exporting theory_annotations compatible with review-mining-stp.

Search arXiv papers by keyword, author, category, or ID. Find related work via Semantic Scholar citations and recommendations.

Use when planning or executing a full data analysis workflow, including schema inspection, data quality audit, data cleaning, EDA, relationship analysis, feature engineering, modeling, evaluation, and report generation. Trigger on requests like 資料分析流程, EDA 到建模, 數據分析規劃, 分析報告產出, or end-to-end analytics workflow.

Use when the user needs stock screening, portfolio risk review, dividend portfolio design, pre-earnings analysis, industry competition analysis, DCF valuation, technical analysis, or stock trend/anomaly detection. Trigger on requests like 選股, 投資組合風險, 股息策略, 財報前瞻, DCF 估值, 技術面分析, 產業競爭研究, 趨勢識別, or investment research prompts.

Use when the user wants to process a backlog of dev tickets end-to-end with a known project / branch / merge target / task tracker / design reference workflow, but without committing to any specific platform (Jira / Linear / GitHub Issues / Notion / Figma / Zeplin / etc.). Trigger on requests like 「幫我跑任務」、「照我的流程把 backlog 做完」、「review failed 的全部處理掉」、「用我那套開發流程做事」、"work through my task list", "run my dev loop", "process my tickets end-to-end". Always asks the user up-front for project root, working branch, merge target, task list location, design reference location, and how to mark a task done — then loops through tickets implementing → verifying → committing → opening PR → merging → syncing → updating the tracker. Never bake in platform-specific API calls; figure those out at run time from the user's answers.

Inspect a diff before merging. Scope drift check, critical code review, specialist parallel dispatch, adversarial review. Triggers on: 幫我看 diff, 看一下改了什麼, review 一下, PR review, diff review, 幫我看看這個, code review, 看看有沒有問題

Design architecture and lock technical decisions for a feature. Produces architecture diagrams, error maps, and convergence artifacts (delivery-plan.md, AGENTS.md, OpenSpec proposals). Triggers on: 技術方案, 架構設計, 怎麼切模組, eng planning, architecture design, 技術規劃, 這個功能技術上怎麼做, design review, UI review, AI slop scan

Plan a feature before building. Brainstorm the idea, challenge scope, and lock decisions one at a time. Triggers on: 幫我想這個功能, 幫我規劃, 這個怎麼做, 幫我想一下, feature planning, planning mode, 規劃一下, 做這個功能之前先想清楚

Systematic debugger. Root cause before any fix. Iron Law: no fixes without investigation. Hypothesis-driven ??form 3-5 hypotheses, test them cheaply before touching code. Stop after 3 failed fix attempts and escalate. Use when stuck on a bug.

Create, configure, and edit OpenClaw agents from a host machine (locally or over SSH). Use this whenever the user wants to add a new OpenClaw agent, build an agent for a business or organization (customer service, internal helpdesk, booking bot, etc.), scaffold agent workspace files (SOUL.md / AGENTS.md / IDENTITY.md / USER.md), wire channels (LINE / WhatsApp / Discord / Telegram / Slack and others) and bindings, set up an agent knowledge base (RAG), harden an agent against prompt injection and identity confusion, or edit / reconfigure / retune an existing OpenClaw agent. Trigger on any mention of OpenClaw agents, openclaw.json, agent bindings, agent workspace, "add an agent", "build a bot", "編一個 agent", "做一個客服 bot" — and also whenever ~/.openclaw or an openclaw.json file is present in the working host even if OpenClaw is not named.

Operate, configure, and debug a running OpenClaw installation from a host machine (locally or over SSH). Covers the whole runtime surface **except** building agent personalities/workspaces — channels, gateway, daemon/service, config (`openclaw.json`), models + auth, cron, hooks, plugins, sandbox, sessions, security audit, doctor, reset, uninstall. Also the go-to runbook for diagnosing OpenClaw issues: gateway probes, channel status, doctor, security audit, logs, sessions. Trigger on any mention of `openclaw` CLI commands, `~/.openclaw`, `openclaw.json`, "restart gateway", "add a channel", "check why my bot isn't responding", "openclaw doctor", "channels status", "ssh into the OpenClaw box", and similar — also whenever a host has `~/.openclaw/openclaw.json` and OpenClaw work is implied even if not named. Hand off to **openclaw-agent-builder** for any work that creates or edits an agent's workspace files (SOUL.md / AGENTS.md / IDENTITY.md / USER.md / RAG knowledge / `agents.list[]` entries / `bindings[]`).

Build services on PostgREST-compatible BaaS (Supabase / InsForge). RLS, Auth, PostgREST queries, MCP settings. Triggers on: 用 Supabase / InsForge 做後端, 設計 RLS policy, 寫 PostgREST query, 處理 BaaS Auth, 設定 MCP, 推 Supabase production, 自架 Supabase, InsForge

Capture and retrieve project learnings. Record pitfalls, patterns, and preferences from real work. Triggers on: 記一下, 之前踩過什麼雷, 學到什麼, 有什麼教訓, remember this, lesson learned, project learnings, 記錄一下, 之前遇過的, 把這個記起來

在專案中建立並維護 Zeabur 部署規範，會在專案根目錄的 AGENTS.md 寫入一段部署約束（Zeabur 只支援 Dockerfile、不支援 docker-compose；compose 僅供本地開發測試）。僅在使用者明確指定專案要部署到 Zeabur 時才使用 — 例如使用者明說「這個專案要部署到 Zeabur」「要上 Zeabur」「設定 Zeabur 部署」，或直接呼叫此 skill。不要在僅僅提到 Docker、寫 Dockerfile、討論部署平台、或泛泛談到 Zeabur 時觸發；必須有明確的「部署到 Zeabur」意圖。Only trigger when the user explicitly states the project deploys to Zeabur — do not infer from Docker usage or general deployment talk.

Prepare and ship a feature branch. Sync, test, open PR. Triggers on: 可以上了嗎, 準備上線, 開 PR, ship this, deploy, release, 上線, merge, push this, 準備好了嗎, ready to go

Run tests, find failures, fix them, verify. Diff-aware: tests the routes/pages your changes affect. Triggers on: 跑測試, 測一下, 有沒有壞掉, run tests, test this, 跑一下 tests, 測試, check if broken, 有沒有問題

Generate hand-drawn style Excalidraw diagrams from text content. Supports three output modes - Obsidian (.md), Standard (.excalidraw), and Animated (.excalidraw with animation order). Choose this over Obsidian Canvas when the user wants an Excalidraw hand-drawn whiteboard style. For standard Mermaid diagrams (flowcharts, sequence diagrams, etc.), use the mermaid-visualizer skill instead. Triggers on "Excalidraw", "畫圖", "Excalidraw流程圖", "Excalidraw心智圖", "標準Excalidraw", "standard excalidraw", "Excalidraw動畫", "動畫圖".

Transform text content into professional Mermaid diagrams for presentations and documentation. Use when users ask to visualize concepts, create flowcharts, or make diagrams from text. For hand-drawn whiteboard style, use excalidraw-diagram. For interactive Obsidian Canvas with spatial organization, use obsidian-canvas-creator. Supports process flows, system architectures, comparisons, mindmaps, and more with built-in syntax error prevention.

Create native Obsidian Canvas files (.canvas) from text content, supporting both MindMap and freeform layouts. Choose this over Excalidraw when the user specifically wants an interactive Obsidian Canvas or spatial organization of notes and cards. For hand-drawn whiteboard style, use excalidraw-diagram. For standard flowcharts/sequence diagrams, use mermaid-visualizer. Triggers on "Canvas", "Obsidian Canvas", "畫Canvas", "Canvas心智圖", "Canvas佈局".