用配對 CSV 核對原始值與報告值、單位、期間及容許誤差。Compare explicitly paired source/report numbers, units and periods from CSV. Readable Python download; no model required.
交件前，找出漏檔、錯版與審核回條不一致。需要檔案與已接受的清單；Python 本機執行，不需模型。Catch missing files, wrong versions and mismatched review receipts. Local Python download; accepted manifest required. No model needed.
Turn pasted AI evaluation cases into an evidence-bounded triage brief that separates expected and observed behavior, hypotheses, and missing checks without claiming a root cause or fix.
1日あたりの投稿枠、補充の頻度と1回あたりの生産数、現在の在庫数を貼るだけで、週の収支とあと何日で枯れるか、週あたり何本足りないかを式の形で返します。休日だけの増枠が計算から漏れる、1回のバッチで作れる本数に上限がある、補充は成功しているのに量が足りない、作成済みと未使用を分けて数えていないなど9項目を検査し、安全在庫の決め方まで返します。
We built Autowin by putting models in a loop: discover upgrades → test them → judge them with an assembly of adversarial sub-agents — and we ran that very process on itself, iterating the kit until each piece held up under its own review. The result: from one loosely-defined need, you get back work that has been built, tested, and scored by a panel of judge sub-agents — with closure authority kept outside the model (deterministic checks + you), so an "it's done" is never just taken on faith.
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。
Rank your supplied topics with explicit scores, reasons and source references. Python/Skill download; no live news feed, model credits or popularity prediction. 依明確分數、理由與來源排列候選題目。Python／Skill 下載包，不含即時新聞、模型額度或人氣預測。
앞두고 있는 결정을 붙여넣으면, "6개월 뒤 이미 실패했다"고 가정하고 실패 원인을 역설계해 위험 순위·조기경보 신호·사전 가드까지 지도로 만들어줍니다. 결정은 대신 내려주지 않습니다.