Development

死代碼清理與整合專家。用於移除未使用的程式碼、重複程式碼和重構。執行分析工具 (knip, depcheck, ts-prune) 識別死代碼並安全移除。

前端品質閘門。自動檢查 SEO、無障礙、效能模式、UI/UX 基線。編輯 .tsx/.css 後自動觸發，push 前強制執行。修復找到的問題，不做架構變更。

Use when planning new features, starting spec-driven development workflows, or when the user mentions OpenSpec, SDD, or spec-driven development. Also triggers when user asks about feature specifications, change proposals, or wants to organize development workflow.

需求不明時的意圖萃取訪談：一次一題、每題附上自己的猜測、聽出「真正想要 vs 覺得應該要」，直到能預測使用者反應（約 95% 信心）才動工。適用：需求缺少對象 / 動機 / 成功標準 / 約束，或使用者點名「訪談我」「先確認一下」「我們確定嗎」。明確自足的指示、純資訊查詢、機械性操作不適用。

FastAPI + SQLModel + Neon PostgreSQL 後端開發最佳實踐。適用於 Render 部署的 API 服務。涵蓋路由設計、資料庫操作、錯誤處理、認證授權。

對非平凡決策啟動新鮮 context 對抗審查（找碴不背書），在修正還便宜的時候抓出錯誤方向。適用：高風險改動（production、資安敏感邏輯、不可逆操作）、不熟的程式碼、要宣稱「這樣是安全的 / 可行的」之前。機械性操作與一行修改不適用。

CIS (Corporate Identity System) 設計規範建立與維護。前端專案的色彩、字型、間距、元件用途統一定義。新專案建立時或前端修改時自動觸發，確保設計一致性。

建構與 TypeScript 錯誤修復專家。當建構失敗或出現類型錯誤時主動使用。僅修復建構/類型錯誤，最小化 diff，不做架構修改。專注於快速讓建構通過。

架構文件稽核。比對 CLAUDE.md 記錄與實際程式碼結構，找出差異並建議更新。大型 refactor 完成後或 OpenSpec 歸檔時手動執行。

Angular 21 + PrimeNG 企業應用開發最佳實踐。適用於 MES、ERP 等企業級應用開發。涵蓋元件設計、狀態管理、效能優化、PrimeNG 元件使用規範。

Full browser UAT for web apps — Playwright testing with console/network error capture, accessibility checks, i18n validation, and bug triage. Use when running screen-by-screen UAT or testing specific features in any web or hybrid app (React, Vue, Angular, Ionic, Next.js, etc).

Diagnose, fix, and prepare Python projects for Vercel deployments. Use for FastAPI, Flask, Django, ASGI/WSGI, Python entrypoints, pyproject.toml, requirements.txt, uv, .python-version, bundle size, Python runtime errors, or Vercel's Python SDK/runtime APIs. Use this skill whenever the user mentions Python and Vercel in the same context, even if they don't explicitly ask for deployment help.

Analyze Vercel billing data, identify primary cost drivers, map them to your code, and provide scored optimization recommendations. Use when asked to optimize Vercel costs, reduce Vercel bill, analyze Vercel usage, or improve Vercel spending efficiency.

Searches X/Twitter for real-time perspectives, dev discussions, product feedback, breaking news, and expert opinions using the X API v2. Provides search with engagement sorting, user profiles, thread fetching, watchlists, and result caching. Use when: (1) user says "x research", "search x for", "search twitter for", "what are people saying about", "what's twitter saying", "check x for", "x search", (2) user needs recent X discourse on a topic (library releases, API changes, product launches, ...

Extracts reusable skills from work sessions. Use when: (1) a non-obvious problem was solved worth preserving, (2) a pattern was discovered that would help future sessions, (3) a workaround or debugging technique needs documentation. Manual invocation only via /skill-extractor command - no automatic triggers or hooks.

- Generating PDF documents (reports, invoices, resumes, forms, certificates) - Creating PDFs that need complex layouts (multi-column, grids, cards) - Building PDFs with inline SVG graphics, charts, or icons - Producing documents with custom Google Fonts or emoji - Any PDF task where flexbox layout is easier than absolute coordinate math

Reference for writing and editing agent skills well — the vocabulary and principles that make a skill predictable. Consult when authoring, reviewing, or pruning a SKILL.md.

Teaches the user a new skill or concept over multiple sessions, using the current directory as a stateful teaching workspace with lessons, learning records, and reference materials.

Compacts the current conversation into a handoff document so a fresh agent can continue the work in a new session.

Interviews the user relentlessly about a plan, decision, or idea until every branch of the decision tree is resolved. Use when the user wants to stress-test their thinking, sharpen a plan or design before acting, or uses any 'grill' trigger phrase (e.g. "grill me on this").

Implements file-based planning for complex multi-step tasks. Creates task_plan.md, findings.md, and progress.md as persistent working memory. Use when starting tasks requiring >5 tool calls, multi-phase projects, research, or any work where losing track of goals and progress would be costly.

Use when tasks involve creating, editing, analyzing, or formatting spreadsheets (`.xlsx`,

Use when the user asks to inspect Sentry issues or events, summarize recent production errors,

Use when the task requires automating a real browser from the terminal (navigation, form