在 Codex 建立專用專案後，由 Agent 為你建立一份獨立的私人網頁版臺指期貨模擬器。手機可透過 4G／5G 或任何 Wi‑Fi 登入使用，電腦不必開機；Yahoo 行情免券商帳號與 API Key，虛擬帳戶、持倉及成交紀錄保存在你自己的私人 Codex Sites 空間。支援大臺、小臺、微臺、多週期 K 線、MA、BOLL、KD、模擬委託、保證金、交易成本與損益，全程不會送出券商真實委託。
Builds a polished Excel financial model for any company. Give it a public company ticker (data pulled automatically from Yahoo Finance and SEC EDGAR) or a private company spreadsheet, and it outputs a fully formatted .xlsx with a 3-statement model (Income Statement, Balance Sheet, Cash Flow) plus DCF valuation — 5 years historical, 5 years projected, WACC build-up, and a sensitivity table.
This tool performs compliance red-line scanning, liability gap analysis, unit economics estimation, and three-scenario stress testing for business models involving prepaid cards, referrals, franchising, and platform commissions, delivering actionable remediation and decision-making reports.
掃描你的價格資料、找出候選策略——而且對每一個講真話。它回測上百個經典組合、排成 leaderboard，然後做 hype 工具不敢做的事：照你試的個數幫最佳 Sharpe 打折（Deflated Sharpe）、檢查樣本外還剩多少、估過擬合機率——漂亮的曲線一律當成待驗證的假設、不是能交易的策略。睡前設好讓它整夜掃，掃越久對你越懷疑。自包含 HTML 報告（中／英）、跑在你自己的 CSV 上。
A pitch deck and business document skill for creating investor presentations, one-pagers, executive summaries, company profiles, and partnership proposals. Produces structured slide-by-slide content, speaker notes, and document copy for startups, SMEs, and consultants seeking funding or partnerships.
把 AI SaaS 產品構想整理成可落地的 MVP 架構藍圖，包含功能範圍、服務拆分、資料模型、API 與非同步流程、權限與資安、部署、成本估算、風險清單，以及 AI 時代的實際開發時程規劃。
Creates truthful, limit-checked, ready-to-paste answers for accelerator, incubator, startup program, pitch competition, and founder applications.
Rebuild a spreadsheet of hardcoded numbers — the kind that breaks when you change one cell and that nobody trusts enough to take to a board or VC — into a driver-based, three-statement linked model you can actually stress. Scope by purpose, build a driver tree (price × volume × retention, not a growth rate from thin air), link P&L → Balance Sheet → Cash Flow, centralize inputs in one assumption layer, run base/bull/bear scenarios with sensitivity, and sanity-check against benchmarks.