bible-fact-check

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Bible Fact Check

圣经事实检查

Systematically audit biblical reference content through 10 quality checks. Report issues with specific line numbers. Never auto-fix — only report.

Checks 1-5 apply to ALL content types (URLs, pasted text, files). Checks 6-10 apply to fun-facts reference files only — skip entirely for URLs and pasted text (do not print N/A lines for them).

通过10项质量检查系统性审核圣经参考内容。报告问题时需标注具体行号。绝不自动修正——仅作报告。

检查1-5适用于所有内容类型（URL、粘贴文本、文件）。检查6-10仅适用于趣味知识参考文件——针对URL和粘贴文本需完全跳过（无需打印N/A行）。

Input Sources (pick one)

输入来源（选其一）

bible-buddy reference file — e.g.,

/review-bible-buddy fun-facts.md

→ reads from

bible-buddy/references/

Any file path — e.g.,
```
/review-bible-buddy /path/to/my-notes.md
```
→ reads the specified file
URL — e.g.,
```
/review-bible-buddy https://example.com/article
```
→ Do NOT use WebFetch — many news sites block it with 403. Run
```
uv run --directory {BIBLE_BUDDY} scripts/fetch_url.py "<URL>"
```
to extract article text. The script auto-falls back to patchright (headless Chromium) when urllib gets 403.
Pasted text — user pastes content directly in the conversation → review the pasted text
No argument — use AskUserQuestion to ask the user whether they want to paste a URL or input text. Do NOT default to any file.

bible-buddy参考文件 — 示例：

/review-bible-buddy fun-facts.md

→ 读取自

bible-buddy/references/

任意文件路径 — 示例：
```
/review-bible-buddy /path/to/my-notes.md
```
→ 读取指定文件
URL — 示例：
```
/review-bible-buddy https://example.com/article
```
→ 请勿使用WebFetch——许多新闻网站会以403状态码拦截。运行
```
uv run --directory {BIBLE_BUDDY} scripts/fetch_url.py "<URL>"
```
提取文章文本。当urllib返回403时，脚本会自动 fallback 到patchright（无头Chromium）。
粘贴文本 — 用户直接在对话中粘贴内容 → 审核粘贴的文本
无参数 — 使用AskUserQuestion询问用户是要粘贴URL还是输入文本。请勿默认使用任何文件。

Path Resolution

路径解析

Resolve

{BIBLE_BUDDY}

before anything else. Check in order:

Project-level:
```
.claude/skills/bible-buddy/
```
(relative to repo root)
User-level:
```
~/.claude/skills/bible-buddy/
```

Use the first path where

references/

directory exists. If neither exists, stop immediately and tell the user: 「bible-fact-check 需要 bible-buddy skill。請先安裝：

專案層級：

npx skills add lancetw/skills/bible-buddy --project

使用者層級：

npx skills add lancetw/skills/bible-buddy

」

All

{BIBLE_BUDDY}

references below use the resolved path.

优先解析

{BIBLE_BUDDY}

。按以下顺序检查：

项目层级：
```
.claude/skills/bible-buddy/
```
（相对于仓库根目录）
用户层级：
```
~/.claude/skills/bible-buddy/
```

使用第一个存在

references/

目录的路径。若两者均不存在，立即停止并告知用户：「bible-fact-check 需要 bible-buddy skill。請先安裝：

專案層級：

npx skills add lancetw/skills/bible-buddy --project

使用者層級：

npx skills add lancetw/skills/bible-buddy

」

以下所有

{BIBLE_BUDDY}

引用均使用解析后的路径。

Prerequisites

前置条件

Run dependency setup (one-time):

bash

uv sync --directory {BIBLE_BUDDY} && uv run --directory {BIBLE_BUDDY} patchright install chromium

运行依赖设置（一次性操作）：

bash

uv sync --directory {BIBLE_BUDDY} && uv run --directory {BIBLE_BUDDY} patchright install chromium

How to Run

运行步骤

Determine the input source
Resolve
{BIBLE_BUDDY}
path (see Path Resolution above)
Read / receive the content
- For
```
fun-facts.md
```
  : Do NOT use Read (exceeds 10K token limit). Use Grep:
```
Grep("^- ", path="{BIBLE_BUDDY}/references/fun-facts.md", output_mode="content", head_limit=0)
```
  → returns all ~134 facts with line numbers in one call

Load reference files as checking criteria (read on demand, not all at once):

```
{BIBLE_BUDDY}/references/anachronism-timeline.md
```
→ Read entire file (small, 43 lines) → for checks 3, 4

{BIBLE_BUDDY}/references/commonly-misread-passages.md

→ Do NOT Read entire file (44KB wide table). Two-step lookup:

Load index:

bash: grep '|' {BIBLE_BUDDY}/references/commonly-misread-passages.md | awk -F'|' '{print $2}' | sed 's/^ *//;s/ *$//' | grep -v '^-' | grep -v '^Scripture'

→ gives all 79 scripture references (~2KB)

For each relevant passage, Grep the full row:

Grep("Isaiah 7:14", path="{BIBLE_BUDDY}/references/commonly-misread-passages.md")

→ for checks 2, 4, 7, 8

```
{BIBLE_BUDDY}/references/yeshua-hermeneutics.md
```
→ Read entire file (small, 122 lines) → for check 8

Run checks sequentially:
- URL or pasted text: run checks 1-5 only
- fun-facts reference file: run all 10 checks
For each check, report findings with line numbers, or "✅ 通過" if clean
End with a summary table

确定输入来源
解析
{BIBLE_BUDDY}
路径（见上方路径解析）
读取/接收内容
- 针对
```
fun-facts.md
```
  ：请勿使用Read（超出10K令牌限制）。使用Grep：
```
Grep("^- ", path="{BIBLE_BUDDY}/references/fun-facts.md", output_mode="content", head_limit=0)
```
  → 一次调用返回所有约134条带行号的知识点
加载参考文件作为检查标准（按需读取，无需一次性全部加载）：
- ```
{BIBLE_BUDDY}/references/anachronism-timeline.md
```
  → 读取整个文件（较小，43行）→ 用于检查3、4
- ```
{BIBLE_BUDDY}/references/commonly-misread-passages.md
```
  → 请勿读取整个文件（44KB宽表）。分两步查找：
  1. 加载索引：
```
bash: grep '|' {BIBLE_BUDDY}/references/commonly-misread-passages.md | awk -F'|' '{print $2}' | sed 's/^ *//;s/ *$//' | grep -v '^-' | grep -v '^Scripture'
```
    → 返回所有79条经文引用（约2KB）
  2. 针对每个相关段落，使用Grep获取完整行：
```
Grep("Isaiah 7:14", path="{BIBLE_BUDDY}/references/commonly-misread-passages.md")
```
    → 用于检查2、4、7、8
- ```
{BIBLE_BUDDY}/references/yeshua-hermeneutics.md
```
  → 读取整个文件（较小，122行）→ 用于检查8
按顺序运行检查：
- URL或粘贴文本：仅运行检查1-5
- 趣味知识参考文件：运行全部10项检查
每项检查需报告发现的问题并标注行号，若无问题则标注「✅ 通過」
结尾附上总结表格

The 10 Checks

10项检查

1. 事實錯誤檢查 (Factual Errors)

Reference files are the minimum, not the ceiling. Also apply your own biblical, historical, and linguistic knowledge to verify ALL factual claims — including date precision (e.g., "1820 年代" vs "1830 年代"), causal direction (did A cause B or vice versa?), and scholarly attribution accuracy.

Verify:

Historical dates (cross-reference
```
anachronism-timeline.md
```
for known dates, AND verify precision — if timeline says "1820s-1830s" but content says only "1820 年代", flag the imprecision)
Person attributions (who did what in which Bible chapter)
Word meanings: if a Hebrew/Greek term appears in
```
commonly-misread-passages.md
```
's "Key Hebrew/Greek Term" column, verify the content uses it correctly. Also check semantic range — flag claims that reduce a word to a single meaning when it has a broader range (e.g., pistis ≠ only "faithfulness")
Scholarly claims (hapax legomenon, manuscript evidence, etc.)
If the content's interpretation of a passage contradicts the "Actual First-Century Jewish Context" column in
```
commonly-misread-passages.md
```
, flag it
Causal claims: verify the direction is correct (e.g., if a Chinese translation was published in 1919, it cannot have "caused" an 1844 doctrine)
Theological teachings without any scripture citation: if the content makes a theological claim (about God, the Spirit, salvation, church practice) without citing scripture, flag as "ungrounded teaching" (無經文根據的教導)

参考文件是最低标准，而非上限。同时运用你自身的圣经、历史和语言学知识验证所有事实性声明——包括日期精度（如「1820 年代」vs「1830 年代」）、因果方向（是A导致B还是反之？）以及学术归因的准确性。

验证内容包括：

历史日期（交叉参考
```
anachronism-timeline.md
```
中的已知日期，同时验证精度——若时间线显示「1820s-1830s」但内容仅写「1820 年代」，需标记精度不足）
人物归因（谁在圣经哪一章做了什么）
词义：若希伯来文/希腊文词汇出现在
```
commonly-misread-passages.md
```
的「Key Hebrew/Greek Term」列，验证内容对其的使用是否正确。同时检查语义范围—— 若内容将一个具有更广语义范围的词汇简化为单一含义（如pistis≠仅指「faithfulness」），需标记
学术声明（hapax legomenon、手稿证据等）
若内容对某段落的解读与
```
commonly-misread-passages.md
```
中「Actual First-Century Jewish Context」列的内容相矛盾，需标记
因果声明：验证方向是否正确（如某中文译本于1919年出版，不可能「导致」1844年的教义）
无经文引用的神学教导：若内容提出神学主张（关于上帝、圣灵、救赎、教会实践）却未引用经文，标记为「無經文根據的教導」

2. 經節編號檢查 (Verse Reference Errors)

Verify every chapter:verse reference matches the actual Bible content cited.

验证每一处章节:節的引用是否与所引用的实际圣经内容匹配。

3. 年代錯置檢查 (Anachronism)

Cross-reference

anachronism-timeline.md

for every doctrine, practice, or concept mentioned. If a doctrine has a known origin date (e.g., altar call = 1830s, TULIP = 1618-1619, spiritual covering = 1970s), verify the content uses the correct date and does not project it back to the first century.

Also check date precision and conflation: if the timeline distinguishes between a doctrine (e.g., Dort's five points, 1618-1619) and a later label (e.g., TULIP acronym, 20th century), flag content that conflates them. If the timeline gives a range (e.g., "1820s-1830s"), flag content that uses only one end of the range without qualification.

Also detect implicit anachronistic terminology even when no verse is cited. Scan the content for terms that match any doctrine/practice in

anachronism-timeline.md

. If a match is found, note the actual origin date as supplementary context — not as a criticism. News articles and practical guides are not expected to cite historical origins. The goal is to give the reader background, not to fault the author for omitting academic footnotes.

Also flag entries that use post-200 CE sources (Mishnah, Talmud, Targums) to definitively explain pre-100 CE texts with language like "這才是...的背景". Later sources CAN illustrate earlier concepts, but the language must be hedged ("可能的背景" not "這才是"). Flag definitive claims only.

交叉参考

anachronism-timeline.md

验证所有提及的教义、实践或概念。若某教义有明确的起源日期（如决志祷告=1830年代， TULIP=1618-1619属灵遮盖=1970年代），验证内容使用的日期是否正确，且未将其追溯至公元一世纪。

同时检查日期的精度和混淆：若时间线区分了某教义（如多特五大要点，1618-1619）和后来的标签（如TULIP缩写，20世纪），需标记混淆两者的内容。若时间线给出范围（如「1820s-1830s」），需标记仅使用范围一端且未加限定的内容。

同时检测隐性的年代错误术语，即使未引用经文。扫描内容中是否有与

anachronism-timeline.md

中任何教义/实践匹配的术语。若找到匹配项，需注明实际起源日期作为补充背景——而非批评。新闻文章和实用指南无需引用历史起源，目的是为读者提供背景信息，而非指责作者省略学术脚注。

同时标记使用公元200年后来源（米示拿、塔木德、他尔根）来明确解释公元100年前文本的条目，如使用「這才是...的背景」这类表述。后期资料可以说明早期概念，但措辞需谨慎（用「可能的背景」而非「這才是」）。仅标记确定性主张。

4. 過度簡化檢查 (Oversimplification)

Find entries that present one scholarly position as definitive when significant debate exists. Look for:

Minority academic views stated as fact
Complex debates reduced to one-sided conclusions
Words like "其實是" or "就是" on debated topics

Cross-reference

commonly-misread-passages.md

: if the content falls into a "Common Misreading" pattern listed there, flag it. Also check

yeshua-hermeneutics.md

: if the content mentions Jesus' teaching methods (parables, arguments from Torah), verify it correctly identifies the method (e.g., kal va-chomer, mashal, remez) rather than oversimplifying as generic "metaphor" or "symbolism".

找出将某一学术立场作为定论，但实际上存在重大争议的条目。注意：

被表述为事实的少数派学术观点
被简化为片面结论的复杂争议
在有争议话题上使用「其實是」或「就是」等词汇

交叉参考

commonly-misread-passages.md

：若内容符合其中列出的「Common Misreading」模式，需标记。同时检查

yeshua-hermeneutics.md

：若内容提及耶稣的教导方法（比喻、从妥拉出发的论证），验证其是否正确识别方法（如kal va-chomer、mashal、remez）而非简化为通用的「隐喻」或「象征」。

5. 數量計算檢查 (Counting Errors)

Verify all numerical claims: word/character counts, verse counts, people counts, ratios, dates.

验证所有数值声明：字数/字符数、节数、人数、比例、日期。

6. 重複檢查 (Duplicates) ⟨fun-facts only⟩

6. 重複檢查 (Duplicates) ⟨仅适用于趣味知识⟩

Skip this check for URLs and pasted text.

Find entries that cover the same verse, topic, or argument. Two entries using the same framework applied to different topics are NOT duplicates. Two entries making the same point about the same verse ARE duplicates.

针对URL和粘贴文本跳过此项检查。

找出涵盖同一经文、主题或论点的条目。将同一框架应用于不同主题的两个条目不算重复。针对同一经文提出相同观点的两个条目才算重复。

7. 稻草人檢查 (Strawman Arguments) ⟨fun-facts only⟩

7. 稻草人檢查 (Strawman Arguments) ⟨仅适用于趣味知识⟩

Skip this check for URLs and pasted text.

Find entries where "不是 X" and X is something nobody in Taiwan would actually believe or confuse. Cross-reference

commonly-misread-passages.md

: if the denied X appears in the "Common Misreading" column, the denial is grounded and valid. If X has no basis in the reference file AND no basis in Taiwan church practice, it is likely a strawman. Test: does the misreading exist in the reference file, or would a Taiwan churchgoer actually think X?

针对URL和粘贴文本跳过此项检查。

找出包含「不是X」且X是台湾无人会真正相信或混淆的内容的条目。交叉参考

commonly-misread-passages.md

：若被否定的X出现在「Common Misreading」列，则该否定是有依据且有效的。若X在参考文件中无依据且不符合台湾教会实践，则很可能是稻草人论证。测试标准：该误读是否存在于参考文件中，或台湾教会会友是否真的会认为X正确？

8. 意味不明檢查 (Unclear Meaning) ⟨fun-facts only⟩

8. 意味不明檢查 (Unclear Meaning) ⟨仅适用于趣味知识⟩

Skip this check for URLs and pasted text.

Find entries that just state a fact or quote a verse without explaining:

What misconception it corrects, OR
What surprising insight it reveals, OR
Why a Taiwan reader should care

Every entry needs a clear "so what?"

If the entry involves a Hebrew/Greek term, check whether it provides the key term from

commonly-misread-passages.md

. Entries that discuss original language without naming the actual word (e.g., saying "原文意思是X" without giving the Hebrew/Greek) are weaker than those that do.

针对URL和粘贴文本跳过此项检查。

找出仅陈述事实或引用经文，但未解释以下内容的条目：

它纠正了什么误解，或
它揭示了什么惊人见解，或
台湾读者为何应该关心

每个条目都需要明确的「那又怎样？」

若条目涉及希伯来文/希腊文词汇，检查是否提供了

commonly-misread-passages.md

中的关键词汇。讨论原文语言却未提及实际词汇的条目（如仅说「原文意思是X」却未给出希伯来文/希腊文）不如明确给出词汇的条目严谨。

9. 字數控制檢查 (Verbosity) ⟨fun-facts only⟩

9. 字數控制檢查 (Verbosity) ⟨仅适用于趣味知识⟩

Skip this check for URLs and pasted text.

Compare each entry's character count to the median of the content. Flag entries significantly longer than the median (roughly 2x or more).

针对URL和粘贴文本跳过此项检查。

将每个条目的字符数与内容的中位数进行比较。标记明显长于中位数的条目（约为中位数的2倍或更多）。

10. 趣味門檻檢查 (Interest Threshold) ⟨fun-facts only⟩

10. 趣味門檻檢查 (Interest Threshold) ⟨仅适用于趣味知识⟩

Skip this check for URLs and pasted text.

Flag entries that are too academic, too niche, or lack relevance for Taiwan readers. Test questions:

Would a Taiwan Sunday school teacher find this interesting?
Does it require specialized knowledge to appreciate?
Is it relevant to what Taiwan churches actually teach or encounter?

Use

commonly-misread-passages.md

as a relevance guide: passages listed there with test cases from folk-religion, marketplace-theology, or charismatic-theology are highly relevant to Taiwan. Passages tied only to foreign contexts (e.g., Hindu theology, Vajrayana) are less relevant unless Taiwan has that community.

针对URL和粘贴文本跳过此项检查。

标记过于学术、过于小众或与台湾读者无关的条目。测试问题：

台湾主日学教师会对此感兴趣吗？
它需要专业知识才能理解吗？
它与台湾教会实际教导或遇到的内容相关吗？

以

commonly-misread-passages.md

作为相关性指南：其中列出的涉及民间宗教、市场神学或灵恩神学测试案例的经文与台湾高度相关。仅与外国背景相关的经文（如印度教神学、密宗）相关性较低，除非台湾存在相关群体。

General Sweep (after completing all 10 checks)

全面扫查（完成所有10项检查后）

Re-read the entire content one more time with fresh eyes. Use your own biblical, historical, and linguistic knowledge to catch anything the 10 structured checks missed. Common things to look for:

Pre-existing antecedents that the content ignores (e.g., "Augustine invented original sin" ignores 4 Ezra, 2 Baruch)
Conflation of related but distinct concepts (e.g., Colwell's rule vs Harner's qualitative argument)
Translation history gaps (e.g., claiming CUV 1919 originated a connection when Morrison 1823 already used it)
Counter-examples the content doesn't address (e.g., Joel 1:8 for betulah)

Report any additional findings under a separate

### 補充：通則檢查

section before the summary. If nothing additional is found, omit this section.

以全新视角重新通读整个内容。运用你自身的圣经、历史和语言学知识捕捉10项结构化检查遗漏的内容。常见需注意的点：

内容忽略的先前先例（如「奥古斯丁发明了原罪」忽略了《以斯拉四书》、《巴录二书》）
相关但不同概念的混淆（如Colwell法则vs Harner定性论证）
翻译历史的空白（如声称和合本1919年首次建立某关联，但马礼逊1823年译本已使用该关联）
内容未提及的反例（如关于betulah的《约珥书》1:8）

在总结前单独添加

### 補充：通則檢查

部分报告任何额外发现。若无额外发现，可省略此部分。

Output Format

输出格式

Environment detection:

Claude Code →

uv run --directory {BIBLE_BUDDY} scripts/detect_desktop.py bible-fact-check

→ save to returned path

Cowork / Claude.ai web → Do NOT save. Tell user: "你可以複製回應內容存檔，或在 Claude.ai 中使用 Artifact 功能下載。"

Filename:

YYYYMMDD_HHmm_<slug>.md

(URL → domain+path slug, pasted →

pasted_text

; slug 一律用

分隔)

When reviewing a URL or pasted text (not a bible-buddy reference file):

Detect Desktop path using the script above
Write the full report to
```
<detected-path>/YYYYMMDD_HHmm_<slug>.md
```
Show only the summary table in the conversation, with the file path

When reviewing a bible-buddy reference file: output the full report in the conversation as before.

环境检测：

Claude Code →

uv run --directory {BIBLE_BUDDY} scripts/detect_desktop.py bible-fact-check

→ 保存至返回路径

Cowork / Claude.ai网页端 → 请勿保存。告知用户：「你可以複製回應內容存檔，或在 Claude.ai 中使用 Artifact 功能下載。」

文件名：

YYYYMMDD_HHmm_<slug>.md

（URL→域名+路径slug，粘贴文本→

pasted_text

；slug一律用

分隔）

当审核URL或粘贴文本（非bible-buddy参考文件）时：

使用上述脚本检测桌面路径
将完整报告写入
```
<detected-path>/YYYYMMDD_HHmm_<slug>.md
```
在对话中仅显示总结表格，并附上文件路径

当审核bible-buddy参考文件时：如前所述在对话中输出完整报告。

Report structure

报告结构

undefined

undefined

審查：[filename or URL or "使用者提供的內容"]

審查：[文件名或URL或「使用者提供的內容」]

來源：[URL if applicable] 日期：[YYYY-MM-DD HH:mm:ss]


For each check that has findings, use a table:

來源：[适用时填写URL] 日期：[YYYY-MM-DD HH:mm:ss]


针对每个有发现的检查，使用表格：

N. 檢查名稱

#	原文摘錄	問題說明
1	「...」	簡短說明
2	「...」	簡短說明


Checks with no findings: `### N. 檢查名稱 — ✅ 通過` (one line, no table)

End with a summary that **explains** the issues, not just counts them:

#	原文摘錄	問題說明
1	「...」	簡短說明
2	「...」	簡短說明


无发现的检查：`### N. 檢查名稱 — ✅ 通過`（单行，无需表格）

结尾附上总结，**解释**问题而非仅统计数量：

總結

檢查	結果
1. 事實錯誤	2 項
2. 經節編號	✅
...	...

檢查	結果
1. 事實錯誤	2 項
2. 經節編號	✅
...	...

主要問題說明

[問題類型] — 2-3 句白話解釋，這些問題為什麼重要、讀者該注意什麼。
...

undefined

[問題類型] — 2-3句白话解释，说明这些问题为何重要、读者该注意什么。
...

undefined

Readability Rules

可读性规则

Reports should be easy to read for a Taiwan churchgoer, not just scholars:

全中文為主 — 希臘文/希伯來文只在必要時附註，不要大段英文引用
短段落 — 每個發現 2-3 句話就好，不要寫論文
白話 — 用台灣教會能理解的語言，避免學術腔
補充背景 ≠ 批評 — 年代錯置檢查的語氣是「補充資訊」，不是指責作者
Reference file 引用只需標示來源名稱和關鍵事實，不要貼原文

报告需便于台湾教会会友阅读，而非仅面向学者：

以全中文为主 — 希腊文/希伯来文仅在必要时附注，不要大段引用英文
短段落 — 每个发现2-3句话即可，不要写论文
白话 — 使用台湾教会能理解的语言，避免学术腔
补充背景≠批评 — 年代错置检查的语气应为「补充资讯」，而非指责作者
引用参考文件只需标示来源名称和关键事实，不要粘贴原文

Important Notes

重要注意事项

Use 上主 (not 耶和華 or YHWH) when referring to the divine name in reports
Do NOT auto-fix anything. Report only.
Be specific: include line numbers and quote the problematic text
When a finding is based on a reference file, cite it: 「根據 anachronism-timeline.md：altar call 起源為 1830s，非 1820s」
When in doubt, flag it — the user will decide whether to act
Some checks may not apply to all content types (e.g., key-terms lists won't have strawman arguments). Skip with "N/A — 此檢查不適用於本內容"
Reference files: read ONLY the 3 files listed in Step 4 under
```
{BIBLE_BUDDY}/references/
```
(resolved path).

在报告中提及神的名字时使用「上主」（而非「耶和華」或YHWH）
绝不自动修正任何内容。仅作报告。
内容需具体：包含行号并引用有问题的文本
当发现基于参考文件时，需注明来源：「根據 anachronism-timeline.md：altar call 起源為 1830s，非 1820s」
存疑时标记出来——由用户决定是否采取行动
部分检查可能不适用于所有内容类型（如关键词列表不会有稻草人论证）。跳过需标注「N/A — 此檢查不適用於本內容」
参考文件：仅读取步骤4中列出的
```
{BIBLE_BUDDY}/references/
```
（解析后路径）下的3个文件。