fact-checker

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Fact Checker

事实核查工具

Verify factual claims in documents and propose corrections backed by authoritative sources.

通过权威来源验证文档中的事实性声明并提出修正建议。

When to use

使用场景

Trigger when users request:

"Fact-check this document"
"Verify these AI model specifications"
"Check if this information is still accurate"
"Update outdated data in this file"
"Validate the claims in this section"

当用户提出以下请求时触发：

"核查这份文档的事实性"
"验证这些AI模型规格"
"检查这些信息是否仍然准确"
"更新此文件中的过时数据"
"验证本节中的声明"

Workflow

工作流程

Copy this checklist to track progress:

Fact-checking Progress:
- [ ] Step 1: Identify factual claims
- [ ] Step 2: Search authoritative sources
- [ ] Step 3: Compare claims against sources
- [ ] Step 4: Generate correction report
- [ ] Step 5: Apply corrections with user approval

复制以下清单跟踪进度：

Fact-checking Progress:
- [ ] Step 1: Identify factual claims
- [ ] Step 2: Search authoritative sources
- [ ] Step 3: Compare claims against sources
- [ ] Step 4: Generate correction report
- [ ] Step 5: Apply corrections with user approval

Step 1: Identify factual claims

步骤1：识别事实性声明

Scan the document for verifiable statements:

Target claim types:

Technical specifications (context windows, pricing, features)
Version numbers and release dates
Statistical data and metrics
API capabilities and limitations
Benchmark scores and performance data

Skip subjective content:

Opinions and recommendations
Explanatory prose
Tutorial instructions
Architectural discussions

扫描文档，找出可验证的陈述：

目标声明类型：

技术规格（上下文窗口、定价、功能）
版本号和发布日期
统计数据和指标
API能力与限制
基准测试分数和性能数据

跳过主观内容：

意见和建议
解释性文字
教程说明
架构讨论

Step 2: Search authoritative sources

步骤2：搜索权威来源

For each claim, search official sources:

AI models:

Official announcement pages (anthropic.com/news, openai.com/index, blog.google)
API documentation (platform.claude.com/docs, platform.openai.com/docs)
Developer guides and release notes

Technical libraries:

Official documentation sites
GitHub repositories (releases, README)
Package registries (npm, PyPI, crates.io)

General claims:

Academic papers and research
Government statistics
Industry standards bodies

Search strategy:

Use model names + specification (e.g., "Claude Opus 4.5 context window")
Include current year for recent information
Verify from multiple sources when possible

针对每个声明，搜索官方来源：

AI模型：

官方公告页面（anthropic.com/news, openai.com/index, blog.google）
API文档（platform.claude.com/docs, platform.openai.com/docs）
开发者指南和发布说明

技术库：

官方文档站点
GitHub仓库（发布版本、README）
包管理仓库（npm, PyPI, crates.io）

一般声明：

学术论文和研究
政府统计数据
行业标准机构

搜索策略：

使用模型名称+规格（例如："Claude Opus 4.5 context window"）
包含当前年份以获取最新信息
尽可能从多个来源验证

Step 3: Compare claims against sources

步骤3：对比声明与来源信息

Create a comparison table:

Claim in Document	Source Information	Status	Authoritative Source
Claude 3.5 Sonnet: 200K tokens	Claude Sonnet 4.5: 200K tokens	❌ Outdated model name	platform.claude.com/docs
GPT-4o: 128K tokens	GPT-5.2: 400K tokens	❌ Incorrect version & spec	openai.com/index/gpt-5-2

Status codes:

✅ Accurate - claim matches sources
❌ Incorrect - claim contradicts sources
⚠️ Outdated - claim was true but superseded
❓ Unverifiable - no authoritative source found

创建对比表格：

文档中的声明	来源信息	状态	权威来源
Claude 3.5 Sonnet: 200K tokens	Claude Sonnet 4.5: 200K tokens	❌ 模型名称过时	platform.claude.com/docs
GPT-4o: 128K tokens	GPT-5.2: 400K tokens	❌ 版本和规格错误	openai.com/index/gpt-5-2

状态代码：

✅ 准确 - 声明与来源一致
❌ 错误 - 声明与来源矛盾
⚠️ 过时 - 声明曾为真但已被取代
❓ 无法验证 - 未找到权威来源

Step 4: Generate correction report

步骤4：生成修正报告

Present findings in structured format:

markdown

undefined

以结构化格式呈现核查结果：

markdown

undefined

Fact-Check Report

事实核查报告

Summary

摘要

Total claims checked: X
Accurate: Y
Issues found: Z

核查的声明总数：X
准确的声明：Y
发现的问题：Z

Issues Requiring Correction

需要修正的问题

Issue 1: Outdated AI Model Reference

问题1：AI模型引用过时

Location: Line 77-80 in docs/file.md Current claim: "Claude 3.5 Sonnet: 200K tokens" Correction: "Claude Sonnet 4.5: 200K tokens" Source: https://platform.claude.com/docs/en/build-with-claude/context-windows Rationale: Claude 3.5 Sonnet has been superseded by Claude Sonnet 4.5 (released Sept 2025)

位置： docs/file.md 第77-80行 当前声明： "Claude 3.5 Sonnet: 200K tokens" 修正建议： "Claude Sonnet 4.5: 200K tokens" 来源： https://platform.claude.com/docs/en/build-with-claude/context-windows 理由： Claude 3.5 Sonnet 已被 Claude Sonnet 4.5（2025年9月发布）取代

Issue 2: Incorrect Context Window

问题2：上下文窗口信息错误

Location: Line 79 in docs/file.md Current claim: "GPT-4o: 128K tokens" Correction: "GPT-5.2: 400K tokens" Source: https://openai.com/index/introducing-gpt-5-2/ Rationale: 128K was output limit; context window is 400K. Model also updated to GPT-5.2

undefined

位置： docs/file.md 第79行 当前声明： "GPT-4o: 128K tokens" 修正建议： "GPT-5.2: 400K tokens" 来源： https://openai.com/index/introducing-gpt-5-2/ 理由： 128K是输出限制；上下文窗口为400K。模型也已更新为GPT-5.2

undefined

Step 5: Apply corrections with user approval

步骤5：获得用户批准后应用修正

Before making changes:

Show the correction report to the user
Wait for explicit approval: "Should I apply these corrections?"
Only proceed after confirmation

When applying corrections:

python

undefined

在进行更改之前：

向用户展示修正报告
等待明确批准："我是否应应用这些修正？"
仅在获得确认后继续

应用修正时：

python

undefined

Use Edit tool to update document

使用Edit工具更新文档

Example:

示例：

Edit( file_path="docs/03-写作规范/AI辅助写书方法论.md", old_string="- Claude 3.5 Sonnet: 200K tokens（约 15 万汉字）", new_string="- Claude Sonnet 4.5: 200K tokens（约 15 万汉字）" )


**After corrections:**

1. Verify all edits were applied successfully
2. Note the correction summary (e.g., "Updated 4 claims in section 2.1")
3. Remind user to commit changes


**修正完成后：**

1. 验证所有编辑已成功应用
2. 记录修正摘要（例如："更新了2.1节中的4项声明"）
3. 提醒用户提交更改

Search best practices

搜索最佳实践

Query construction

查询构建

Good queries (specific, current):

"Claude Opus 4.5 context window 2026"
"GPT-5.2 official release announcement"
"Gemini 3 Pro token limit specifications"

Poor queries (vague, generic):

"Claude context"
"AI models"
"Latest version"

优质查询（具体、时效性强）：

"Claude Opus 4.5 context window 2026"
"GPT-5.2 official release announcement"
"Gemini 3 Pro token limit specifications"

劣质查询（模糊、通用）：

"Claude context"
"AI models"
"Latest version"

Source evaluation

来源评估

Prefer official sources:

Product official pages (highest authority)
API documentation
Official blog announcements
GitHub releases (for open source)

Use with caution:

Third-party aggregators (llm-stats.com, etc.) - verify against official sources
Blog posts and articles - cross-reference claims
Social media - only for announcements, verify elsewhere

Avoid:

Outdated documentation
Unofficial wikis without citations
Speculation and rumors

优先选择官方来源：

产品官方页面（权威性最高）
API文档
官方博客公告
GitHub发布版本（针对开源项目）

谨慎使用：

第三方聚合平台（llm-stats.com等）- 需与官方来源验证
博客文章 - 交叉验证声明
社交媒体 - 仅用于公告类信息，需在其他渠道验证

避免使用：

过时文档
无引用的非官方维基
猜测和谣言

Handling ambiguity

处理歧义

When sources conflict:

Prioritize most recent official documentation
Note the discrepancy in the report
Present both sources to the user
Recommend contacting vendor if critical

When no source found:

Mark as ❓ Unverifiable
Suggest alternative phrasing: "According to [Source] as of [Date]..."
Recommend adding qualification: "approximately", "reported as"

当来源存在冲突时：

优先选择最新的官方文档
在报告中注明差异
向用户展示两个来源
若为关键问题，建议联系供应商

当未找到来源时：

标记为 ❓ 无法验证
建议使用替代表述："根据[来源]截至[日期]..."
建议添加限定词："大约"、"据报道"

Special considerations

特殊注意事项

Time-sensitive information

时效性信息

Always include temporal context:

Good corrections:

"截至 2026 年 1 月" (As of January 2026)
"Claude Sonnet 4.5 (released September 2025)"

Poor corrections:

"Latest version" (becomes outdated)
"Current model" (ambiguous timeframe)

始终包含时间上下文：

优质修正表述：

"截至2026年1月"
"Claude Sonnet 4.5（2025年9月发布）"

劣质修正表述：

"最新版本"（会很快过时）
"当前模型"（时间范围模糊）

Numerical precision

数值精度

Match precision to source:

Source says: "approximately 1 million tokens" Write: "1M tokens (approximately)"

Source says: "200,000 token context window" Write: "200K tokens" (exact)

与来源的精度保持一致：

来源表述： "约100万tokens" 修正后： "1M tokens（约）"

来源表述： "200,000 token上下文窗口" 修正后： "200K tokens"（精确值）

Citation format

引用格式

Include citations in corrections:

markdown

> **注**：具体上下文窗口以模型官方文档为准，本书写作时使用 Claude Sonnet 4.5 为主要工具。

Link to sources when possible.

在修正中包含引用：

markdown

> **注**：具体上下文窗口以模型官方文档为准，本书写作时使用 Claude Sonnet 4.5 为主要工具。

尽可能添加来源链接。

Examples

示例

Example 1: Technical specification update

示例1：技术规格更新

User request: "Fact-check the AI model context windows in section 2.1"

Process:

Identify claims: Claude 3.5 Sonnet (200K), GPT-4o (128K), Gemini 1.5 Pro (2M)
Search official docs for current models
Find: Claude Sonnet 4.5, GPT-5.2, Gemini 3 Pro
Generate report showing discrepancies
Apply corrections after approval

用户请求： "核查2.1节中AI模型的上下文窗口"

流程：

识别声明：Claude 3.5 Sonnet（200K）、GPT-4o（128K）、Gemini 1.5 Pro（2M）
搜索官方文档获取当前模型信息
发现：Claude Sonnet 4.5、GPT-5.2、Gemini 3 Pro
生成显示差异的报告
获得批准后应用修正

Example 2: Statistical data verification

示例2：统计数据验证

User request: "Verify the benchmark scores in chapter 5"

Process:

Extract numerical claims
Search for official benchmark publications
Compare reported vs. source values
Flag any discrepancies with source links
Update with verified figures

用户请求： "验证第5章中的基准测试分数"

流程：

提取数值声明
搜索官方基准测试出版物
对比报告值与来源值
标记所有差异并附上来源链接
使用经过验证的数据更新文档

Example 3: Version number validation

示例3：版本号验证

User request: "Check if these library versions are still current"

Process:

List all version numbers mentioned
Check package registries (npm, PyPI, etc.)
Identify outdated versions
Suggest updates with changelog references
Update after user confirms

用户请求： "检查这些库的版本是否仍为最新"

流程：

列出所有提及的版本号
检查包管理仓库（npm、PyPI等）
识别过时版本
建议更新并附上变更日志参考
获得用户确认后更新

Quality checklist

质量检查清单

Limitations

局限性

This skill cannot:

Verify subjective opinions or judgments
Access paywalled or restricted sources
Determine "truth" in disputed claims
Predict future specifications or features

For such cases:

Note the limitation in the report
Suggest qualification language
Recommend user research or expert consultation

本工具无法：

验证主观意见或判断
访问付费墙后的受限来源
判定争议性声明的“真相”
预测未来的规格或功能

针对此类情况：

在报告中注明局限性
建议使用限定性语言
建议用户自行研究或咨询专家