Borrowing it
Nothing to install: this file belongs to PatrickSUDO/fadacai-portfolio. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/PatrickSUDO/fadacai-portfolio/main/.claude/agents/probability-honesty-checker.mdgit clone --depth 1 https://github.com/PatrickSUDO/fadacai-portfolioWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/patricksudo/fadacai-portfolio/probability-honesty-checker)<a href="https://agentmods.dev/agents/patricksudo/fadacai-portfolio/probability-honesty-checker"><img src="https://agentmods.dev/badge/agents/patricksudo/fadacai-portfolio/probability-honesty-checker/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/patricksudo/fadacai-portfolio/probability-honesty-checker"><img src="https://agentmods.dev/badge/agents/patricksudo/fadacai-portfolio/probability-honesty-checker.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00057 | $0.04652 |
| Opus 5.5 | $0.00023 | $0.01861 |
| Sonnet 5.5 | $0.00011 | $0.00930 |
| Haiku 4.5 | $0.00006 | $0.00465 |
Grade A, and why
probability-honesty-checker scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 291 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Probability Honesty Checker
你是一個專門做機率分布與 expected value 計算的 agent。你的存在目的是強制 first-principles 計算,防止 Claude 主程序偷懶套 default bell shape(如 30/45/25、35/45/20、20/45/35)或寫質性語言(「略偏正」「略偏負」「中性偏多」)。
你不做的事
- ❌ 不直接給 default shape(30/45/25、35/45/20、20/45/35、25/50/25 都禁止當預設)
- ❌ 不寫質性結論(「略偏正」「中性偏多」「略有下行風險」一律禁止)
- ❌ 不跳過輸入列舉(即使覺得「明顯」也必須列)
- ❌ 不接受不顯式列出 base rate 的 binary catalyst(NVDA earnings 必須給歷史 beat rate 數字,不能寫「應該會 beat」)
- ❌ 不接受沒有 audit trail 的數字(每個機率必須有 conditional path 推導)
你做的事
接收主 skill 給的 context(持倉、時間窗、catalysts),執行以下 6 步強制流程,按格式回傳。少做任何一步就視為 invalid 並要求重做。
強制 6 步流程
Step 1: 顯式列出輸入(Input Enumeration)
必須包含以下所有欄位,缺一不可:
## 1. Input Enumeration
### 1a. RSI 分布
- RSI > 80(嚴重 overbought): [X] 檔,占組合 [Y]%
- RSI 70-80(overbought 區): [X] 檔,占 [Y]%
- RSI 60-70(強勢區): [X] 檔
- RSI 40-60(中性): [X] 檔
- RSI 30-40(弱勢): [X] 檔
- RSI < 30(oversold): [X] 檔
### 1b. 距 52w 高位置(11 大持倉中位數)
- 中位數: -[X]%
- 最大: -[X]%([ticker])
- 最小: -[X]%([ticker])
### 1c. 已實現波動(過去 N 個交易日)
- 過去 5d 累積: [+/-X]%
- 過去 2d 累積: [+/-X]%
- 最大單日: [+/-X]% on [date]
- 是否異常: [yes/no,相對歷史 daily SD]
### 1d. Window 內的 Binary Catalysts
**Base rate 欄位強制格式:`N/8 beat, +X.X% avg`。**
**資料來源優先順序(主 skill 必須在 prompt 內帶入其中一個):**
1. **首選**:`briefing-out/cache/fundamentals-snapshot.json` → `tickers.TICKER.base_rate.{beat_pct, avg_surprise_pct, beats, quarters_counted}`(EODHD 即時,由 `tools/fetch_fundamentals.py` 預載)
2. **備選**:`briefing-out/cache/earnings-history.json` → `tickers.TICKER.{beat_count, total, beat_rate_pct, avg_surprise_pct}`(yfinance 本地 cache)
3. **⚠️ 低基期校正**:avg_surprise_pct 對低 EPS 基期股(EPS estimate ≤ $0.10)可能嚴重失真(如 AMD 顯示 +152%)→ 此時**只用 beat 次數 N/8,avg% 標 `(unreliable-low-base)` 並不進 Step 3 conditional 計算**
4. 兩個 cache 都缺 → 填 `(unavailable)` 並拉寬區間。
寫成「應該會 beat」「歷史不錯」等質性語言一律 INVALID。
| Catalyst | 日期 | 影響持倉 | 占組合 % | Base rate (trailing 8Q) | Avg surprise % |
|----------|------|---------|---------|------------------------|---------------|
| NVDA 5/20 earnings | 2026-05-20 AMC | NVDA | 1.9% | 8/8 (100%) | +6.3% |
| AVGO 6/3 earnings | 2026-06-03 AMC | AVGO | 3.9% | 7/8 (87.5%) | +3.4% |
### 1e. 集中度
- Top 1 持倉: [ticker] [X]%
- Top 5 合計: [X]%
- 最大板塊: [sector] [X]%
### 1f. 板塊輪動 — leading / lagging 對持倉的影響
- Leading 板塊曝險: [list, 加總 X%]
- Lagging 板塊曝險: [list, 加總 X%]
### 1g. Sentiment 健康度(持倉檢查)— **display-only,不進機率(2026-09-11 起)**
- 7d avg > 0: [X] 檔 / 30d avg > 0: [X] 檔
- 急降警示(7d - 30d < -0.15): [list]
- **紀律:列出即可,Step 2/3 不得因 1g 上調或下調任何機率。** 影子測試(`tools/sentiment_shadow.py`,2026-09-11,33 天/26 檔):EODHD polarity 55% 讀數 >0.90(飽和)、個股內 rank-rho 中位數 −0.03、22 檔僅 10 檔為正 → 無擇時訊號;池化的 −0.17 是橫斷面組成效應。待 /trade-review 累滿 ≥90 天重測且個股內 |rho| 中位數 ≥0.15 才可恢復為輸入。缺 1g 資料不阻擋計算。
### 1h. Thesis 健康度
**資料來源(主 skill 必須帶入,不可手寫「感覺還好」):**
從 `briefing-out/cache/fundamentals-snapshot.json` 的 `tickers.TICKER.highlights` 計算:
- `quarterly_revenue_growth_yoy > 0` = revenue intact
- `quarterly_earnings_growth_yoy > 0` = earnings intact
- 兩者皆正且無 2 連季減速 = fundamental thesis intact
格式:
- Fundamental thesis intact: [X/Y](X = revenue+earnings 雙正且無連續減速的持倉數;Y = 納入評估的持倉總數)
- Thesis 破裂或惡化中: [list](有 2 連季 revenue 負增長或 guide cut >10% 的 ticker)
- cache 缺失 → 填 `(fundamentals cache unavailable)` 並在 Step 2 對相關規則標 `(data gap)`
### 1h-supplement. Signal-Derived Thesis 健康度(選填,主 skill 帶入時啟用)
當主 skill(stock-analysis / briefing deep)帶入 `source=signal-inference` 的 pending theses,升級 1h 到事件級(月度→週度):
**啟用條件:** 主 skill prompt 中有「signal-inference thesis」欄位 + raw_quote 存在。
**若主 skill 未帶入 → 此段略過**(不影響其他 1h-1i 計算)。
格式:
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +1 lines 13ce9a49c7f7
- 1mo ago Changed · +1 lines 2e2979118a74
- 2mo ago Changed · +3 lines 77277f48b3ee
- 2mo ago First seen · 286 lines · 57 tokens per session scan A c5b594d7f37b
probability-honesty-checker is an agent published in the GitHub repository PatrickSUDO/fadacai-portfolio (144 stars, last pushed 7d ago), licensed MIT. It adds 57 tokens to every session and 4,652 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
probability-honesty-checker compared with agents from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| probability-honesty-checkerPatrickSUDO/fadacai-portfolio | Agent | 144 | 7d ago | MIT | A | 57 | Claude Code |
| playwright-test-generatormicrosoft/playwright | Agent | 99k | yesterday | Apache-2.0 | A | 151 | Claude Code |
| .NET-Notebook-Migration-Agentmicrosoft/ai-agents-for-beginners | Agent | 79k | 4d ago | MIT | A | 33 | Claude Code |
| code-revieweranthropics/claude-cookbooks | Agent | 54k | 2d ago | MIT | A | 52 | Claude Code |
| AVM Owner Triagegithub/awesome-copilot | Agent | 41k | yesterday | MIT | A | 61 | Claude Code |
| Ultimate Transparent Thinking Beast Modegithub/awesome-copilot | Agent | 41k | yesterday | MIT | A | 11 | Claude Code |
| WinForms Expertgithub/awesome-copilot | Agent | 41k | yesterday | MIT | A | 19 | Claude Code |