Ev Check by PatrickSUDO — AI Skill for Any LLM

Q: What does the Ev Check AI skill do?

強制 first-principles 機率分布 + EV 計算。用於檢查當前組合在指定時間窗的預期報酬，禁止用 default bell shape 或質性語言。Usage - /ev-check [30d|7d|14d] [optional scenario theme]

Q: Why use Ev Check on TypingMind?

Because you install it once and use it with any model. Ev Check is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

Q: How do I install Ev Check in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/PatrickSUDO/fadacai-portfolio/tree/main/.agents/skills/ev-check. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Q: Which AI models can use Ev Check?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

Q: How many AI models can I use with Ev Check?

As many as you like. As long as a model supports skills, you can use Ev Check with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Q: Is the Ev Check AI skill free?

Yes. It is published on GitHub by PatrickSUDO under the MIT license. You only pay your own AI provider for the tokens you use.

Q: What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

Q: How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

Overview

Publisher	PatrickSUDO
Repository	`fadacai-portfolio`
Skill name	`ev-check`
Stars	144
Forks	12
Bundled files	Instructions only
License	MIT
Links	Skill source Repository

Markdown instructions

A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

Works with any LLM

AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

Self-contained

Everything the model needs lives in the instructions — no extra files to sync.

Open source

Published by PatrickSUDO on GitHub. Read the source before you install it.

Installation

Install the Ev Check AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

Open the app and go to Plugins → Skills.
Choose "Install from GitHub".
Paste the skill folder URL below and confirm.
https://github.com/PatrickSUDO/fadacai-portfolio/tree/main/.agents/skills/ev-check
Enable the skill in any chat where you want it available.

Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills

git clone --depth 1 https://github.com/PatrickSUDO/fadacai-portfolio.git /tmp/fadacai-portfolio
mkdir -p .claude/skills
cp -r /tmp/fadacai-portfolio/.agents/skills/ev-check .claude/skills/ev-check

Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Ev Check in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Ev Check on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Ev Check is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

Loaded only when it is needed

The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

Switch models mid-chat

Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

EV / Probability Distribution Honesty Check

對當前持倉執行嚴謹的機率分布與 expected value 計算。強制 first-principles，不接受偷懶輸出。

此 skill 可獨立呼叫做 ad-hoc check，或被其他 skill（briefing / portfolio-review / stock-analysis / todo）在輸出 Verdict / 機率分布前 mandatory 呼叫。

Arguments

/ev-check → 預設 30 天 horizon
/ev-check 7d / /ev-check 14d / /ev-check 30d → 自選時間窗
/ev-check 30d nvda-bear → 用戶指定情境主題（agent 會以此為主要 catalyst 反推）
/ev-check 30d 比選 ... → 名額比選模式。比選集合不是 briefing 點名的那幾檔，而是 plan.md 候補表整層 L1 全掃（2026-09-21 用戶質問「選 HWM 不選 ATI 什麼理由」，當次只比了 briefing 框的 HWM/CRDO，ATI 漏掉）。作法：先列 L1 每一檔的「進場分支是否成立（價格線/走強線）、分析師覆蓋 N、與空名額同鏈的上下游關係」做一張淘汰表，明寫每檔為何進或不進 EV 計算；只對淘汰表存活者跑 agent。淘汰表要出現在輸出裡，用戶事後問任何一檔都要能指出被淘汰的那一行。落選者同一次撤提名警報（shadow_signals.py record --kind cf-bench-loser 已自動撤該標的 price_above/rolling_high 線；輸出列「已撤警報」欄，見 feedback/alert-hygiene.md）。

Workflow

Step 1: 收集 raw 輸入

呼叫以下 MCP 取數據（可平行）：

mcp__firstrade-server__get_account_position — 持倉
mcp__firstrade-server__get_account_balance — 帳戶總值 + 現金
mcp__technical-mcp__get_batch_indicators(所有持倉, period=3mo) — RSI / momentum / trend
mcp__technical-mcp__get_sector_rotation(period=3mo) — leading / lagging
mcp__yfinance-advanced__get_stock_info(top 11 by MV) — 52w high/low、fundamentals
mcp__fmp-mcp__getEarningsCalendar(today, today+horizon) — binary catalysts in window
（平行 agent）mcp__eodhd-mcp__get_sentiment_trend(top 8 by MV, days=30) — 7d/30d sentiment
Read briefing-out/cache/macro-snapshot.json — macro state（fed_funds / 2s10s / HY OAS / VIX / CPI / regime_tag）；status == "skipped" 或缺失 → 1i 標 unavailable 並註明

Step 2: 整理成 9 項 Input Enumeration

按 probability-honesty-checker agent 的 Step 1 contract，整理：

1a. RSI 分布（每個 bucket 檔數 + % of port）
1b. 距 52w 高位置（中位數、最大、最小）
1c. 已實現波動（5d / 2d / 最大單日）
1d. Binary catalysts table（catalyst / 日期 / 影響持倉 % / base rate）
1e. 集中度（top 1、top 5、最大板塊）
1f. 板塊輪動曝險（leading 持倉 % / lagging 持倉 %）
1g. Sentiment 健康度（display-only，不進機率；2026-09-11 影子測試無擇時訊號）
1h. Thesis 健康度（從 plan.md + 近期新聞）
1i. Macro state（從 macro-snapshot.json：fed_funds + 30d change / 2s10s + regime / hy_oas + regime + pct_1y / vix + regime / cpi_yoy + trend / regime_tag；agent 缺 1i 會回 INVALID INPUT）

不齊全 → 不能進下一步，必須補齊。

Step 3: 呼叫 probability-honesty-checker subagent


Agent(
  subagent_type: "probability-honesty-checker",
  description: "EV check for [horizon]",
  prompt: """
  執行 6 步強制流程計算當前組合 [horizon] 機率分布與 EV。

  時間窗: [horizon]
  情境主題（如有）: [user-specified]

  ## Step 1 輸入資料（8 項齊全）:
  [貼上 Step 2 整理好的資料]

  ## 額外 context:
  [plan.md 摘要 / 用戶提到的特定 catalyst / 最近 N 天的事件]

  請按你的 6 步流程輸出：
  1. Input Enumeration（confirm 我給的齊全）
  2. 形狀反推
  3. Conditional Probabilities
  4. Aggregated Scenario Probabilities
  5. EV Calculation
  6. Self-Audit Checklist
  + 給主 skill 的精簡輸出
  """
)

Step 3.4: EV 怎麼用才不算濫用（2026-09-23 用戶 push back「跟 SGOV 比很不公平」，三條修正）

基準看情境，不是永遠 SGOV。 組合滿編（砍一進一）時，新倉的比較對象是最弱在倉那檔的 EV（機會成本閘門原文），SGOV 只在「另一個選擇真的是現金」（R24 閒置、名額空著）時才是基準。拿 SGOV 判換手決策 = 用錯閘門。
EV 必附誤差帶，差距在帶內不得當裁決依據。 ev_ledger.py stats 2026-09-23：已驗收 15 筆，平均絕對誤差 10.0pp、平均偏差 −8.9pp（歷史偏樂觀）、Brier 技能分數 <0。輸出格式一律「EV −1.7%（±10pp）」；兩個選項 EV 差 <10pp 視為無法區分，裁決改看其他維度（相關度、thesis 成立機率、加權下行、載具）。n≥30 獨立樣本後依實測誤差重訂帶寬。
有出場閘門的部位另算「規則管理版 EV」。 365d 買抱不動的 EV 會把最壞 regime 整段吃進去；實際部位有 R23 線 / thesis gate / R14 後的機械出場，左尾被截斷。做法：對 thesis 破的 regime 把報酬改為「閘門觸發時的截斷損失」（財報 gate 通常 −15%~−20%）重算 Σ，兩版並列，標明哪一版對應用戶實際的持有方式。

Why： NET 9/23 案——報告以「365d EV −9.6% 輸 SGOV 13.5pp」判不建倉，但 ①當時 18/18 滿編、真正對手是 MYRG；②5.7pp（FSLY）的差距在 10pp 誤差內；③用戶的部位有 10/29 硬閘門，截斷後 EV 約 −5%。用戶「感覺就不成立」是對的：原則沒錯（風險資產期望值須高於無風險利率），錯在用錯基準與把 ±10pp 的數字當精準門檻。

Step 3.5: EV ledger 事前登錄（強制，agent 輸出後執行）


python3 tools/ev_ledger.py add --ticker PORTFOLIO --slug <主題>-<horizon> \
  --horizon-days <N> --spot <帳戶即時總值> \
  --p-bull XX --p-base XX --p-bear XX --ev-pct <X.XX> \
  --source ev-check --model <本次模型>

機率/EV 直接抄 agent 輸出；spot 用帳戶即時總值（resolve 時對 research/equity-marks.json 最近標記）。到期由 briefing resolve-due 驗收，校準由 /trade-review 讀 stats。

Step 4: 顯示 agent 完整輸出

不刪減、不簡化、不重寫。直接呈現 agent 的 6 步流程 + 精簡輸出。

主 skill 看到的格式：


# EV Check — [horizon]

[Agent 完整 6 步輸出]

---

## ✅ 給用戶的精簡結論

機率分布：樂觀 X% / 基準 X% / 悲觀 X%
EV ([horizon]) = X.XX%
主導因素：[1 句]

Step 5: 用戶 push back 處理

如果用戶質疑「你真的有算嗎」「這是 default 嗎」：

不要辯解、不要重組原數字
重跑 Step 3（重新呼叫 agent，明確要求 audit checklist 全勾）
如果發現原本確實偷懶（例如 sum 沒到 100%、機率是 default mirror、EV 寫質性語言）→ 老實承認 + 顯示新算

何時被其他 skill 呼叫

以下 skill 在輸出 Verdict / 機率分布 / EV / Quick Take 之前必須呼叫此 skill（或直接 invoke probability-honesty-checker agent）：

/briefing（任何 tier）→ Quick Take 前
/portfolio-review → Section K 第一性檢查前
/stock-analysis → 個股 Verdict 前
/todo → 行動清單第一性檢查前

呼叫方式可選：

走 ev-check skill（完整流程，給用戶看的格式）
直接 invoke probability-honesty-checker agent（內部使用，省一層 wrapping）

Output Format

繁體中文。所有數字 explicit。絕不出現：「略偏正」「略偏負」「中性偏多」「應該會」「不確定性高」等質性語言 — 全部換成數字區間或機率。

失敗模式與防呆

Claude 主程序常見偷懶	此 skill 阻擋方式
套 30/45/25 default	Agent Step 2 強制顯示「形狀規則應用」對照表，不對照不能進 Step 3
寫「略偏負」結論	Agent Step 5 強制顯式 Σ 計算，數字必須出現
跳過 Step 1 直接給機率	Agent 收到不完整 input 回「INVALID INPUT」拒絕計算
Sum ≠ 100%	Agent Step 4 顯式 sum check
中點手動偏移	Agent Step 5 強制 (max+min)/2 算術平均

Frequently asked questions

What does the Ev Check AI skill do?

強制 first-principles 機率分布 + EV 計算。用於檢查當前組合在指定時間窗的預期報酬，禁止用 default bell shape 或質性語言。Usage - /ev-check [30d|7d|14d] [optional scenario theme]

Why use Ev Check on TypingMind?

Because you install it once and use it with any model. Ev Check is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Ev Check in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/PatrickSUDO/fadacai-portfolio/tree/main/.agents/skills/ev-check. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Ev Check?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Ev Check?

As many as you like. As long as a model supports skills, you can use Ev Check with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Ev Check AI skill free?

Yes. It is published on GitHub by PatrickSUDO under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

Overview

Installation

Install in TypingMind

Install in another agent

Use it in TypingMind

Works with any AI model

Skill instructions

EV / Probability Distribution Honesty Check

Arguments

Workflow

Step 1: 收集 raw 輸入

Step 2: 整理成 9 項 Input Enumeration

Step 3: 呼叫 probability-honesty-checker subagent

Step 3.4: EV 怎麼用才不算濫用（2026-09-23 用戶 push back「跟 SGOV 比很不公平」，三條修正）

Step 3.5: EV ledger 事前登錄（強制，agent 輸出後執行）

Step 4: 顯示 agent 完整輸出

Step 5: 用戶 push back 處理

何時被其他 skill 呼叫

Output Format

失敗模式與防呆

Frequently asked questions

Related AI Skills

Set up your own AI workspace now