review-ai-writing
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseAI Writing Detection for Developer Text
面向开发者文本的AI写作检测
Detect patterns characteristic of AI-generated text in developer artifacts. These patterns reduce trust, add noise, and obscure meaning.
检测开发者制品中AI生成文本的典型模式。这些模式会降低可信度、增加冗余并模糊语义。
Pattern Categories
模式分类
| Category | Reference | Key Signals |
|---|---|---|
| Content | | Promotional language, vague authority, formulaic structure, synthetic openers |
| Vocabulary | | AI word tiers, copula avoidance, rhetorical devices, synonym cycling, commit inflation |
| Formatting | | Boldface overuse, emoji decoration, heading restatement |
| Communication | | Chat leaks, cutoff disclaimers, sycophantic tone, apologetic errors |
| Filler | | Filler phrases, excessive hedging, generic conclusions |
| Code Docs | | Tautological docstrings, narrating obvious code, "This noun verbs", exhaustive enumeration |
| 分类 | 参考文档 | 关键信号 |
|---|---|---|
| 内容 | | 宣传性语言、模糊的权威性、公式化结构、合成式开篇 |
| 词汇 | | AI词汇层级、避免系动词、修辞手段、同义词循环、提交信息冗余 |
| 格式 | | 过度使用粗体、表情符号装饰、标题重复表述 |
| 沟通风格 | | 聊天痕迹、截断声明、谄媚语气、致歉式错误表述 |
| 冗余内容 | | 冗余短语、过度模糊表述、通用结论 |
| 代码文档 | | 同义反复的文档字符串、赘述明显代码、"This noun verbs"结构、详尽枚举 |
Scope
检测范围
Scan these artifact types:
| Artifact | File Patterns | Notes |
|---|---|---|
| Markdown docs | | READMEs, guides, changelogs |
| Docstrings | | Language-specific docstring formats |
| Code comments | Same as docstrings | Inline and block comments |
| Commit messages | | Use synthetic path |
| PR descriptions | GitHub PR body | Use synthetic path |
扫描以下制品类型：
| 制品类型 | 文件模式 | 说明 |
|---|---|---|
| Markdown文档 | | README、指南、变更日志 |
| 文档字符串 | | 各语言特定的文档字符串格式 |
| 代码注释 | 与文档字符串相同 | 行内注释和块注释 |
| 提交信息 | | 使用合成路径 |
| PR描述 | GitHub PR正文 | 使用合成路径 |
What NOT to Scan
不扫描的内容
- Generated code (lock files, compiled output, vendor directories)
- Third-party content (copied license text, vendored docs)
- Code itself (variable names, string literals used programmatically)
- Test fixtures and mock data
- 生成代码（锁文件、编译输出、供应商目录）
- 第三方内容（复制的许可证文本、供应商提供的文档）
- 代码本身（变量名、程序中使用的字符串字面量）
- 测试夹具与模拟数据
Detection Rules
检测规则
High-Confidence Signals (Always Flag)
高置信度信号（始终标记）
These patterns are strong indicators of AI-generated text:
- Chat leaks — "Certainly!", "I'd be happy to", "Great question!", "Here's" as sentence opener
- Cutoff disclaimers — "As of my last update", "I cannot guarantee"
- High-signal AI vocabulary — delve, utilize (as "use"), whilst, harnessing, paradigm, synergy
- "This noun verbs" in docstrings — "This function calculates", "This method returns"
- Synthetic openers — "In today's fast-paced", "In the world of"
- Sycophantic code comments — "Excellent approach!", "Great implementation!"
这些模式是AI生成文本的强烈指标：
- 聊天痕迹 —— "Certainly!"、"I'd be happy to"、"Great question!"、以"Here's"开头的句子
- 截断声明 —— "As of my last update"、"I cannot guarantee"
- 高信号AI词汇 —— delve、utilize（用作"use"）、whilst、harnessing、paradigm、synergy
- 文档字符串中的"This noun verbs"结构 —— "This function calculates"、"This method returns"
- 合成式开篇 —— "In today's fast-paced"、"In the world of"
- 谄媚的代码注释 —— "Excellent approach!"、"Great implementation!"
Medium-Confidence Signals (Flag in Context)
中置信度信号（结合上下文标记）
Flag when 2+ appear together or pattern is repeated:
- Low-signal AI vocabulary clusters — 3+ words from the low-signal list in one section
- Formulaic structure — Rigid intro-body-conclusion in a README section
- Heading restatement — First sentence after heading restates the heading
- Excessive hedging — "might potentially", "could possibly", "it seems like it may"
- Synonym cycling — Same concept called different names within one section
- Boldface overuse — More than 30% of sentences contain bold text
当出现2个及以上或模式重复时标记：
- 低信号AI词汇集群 —— 同一区域出现3个及以上低信号列表中的词汇
- 公式化结构 —— README章节中严格遵循引言-正文-结论结构
- 标题重复表述 —— 标题后的第一句重复标题内容
- 过度模糊表述 —— "might potentially"、"could possibly"、"it seems like it may"
- 同义词循环 —— 同一概念在一个章节内被冠以不同名称
- 过度使用粗体 —— 超过30%的句子包含粗体文本
Low-Confidence Signals (Note Only)
低置信度信号（仅作记录）
Mention but don't flag as issues:
- Emoji in technical docs — May be intentional project style
- Filler phrases — Some are common in human writing too
- Generic conclusions — May be appropriate for summary sections
- Commit inflation — Some teams prefer descriptive commits
仅提及但不标记为问题：
- 技术文档中的表情符号 —— 可能是项目的有意风格
- 冗余短语 —— 部分冗余短语在人类写作中也很常见
- 通用结论 —— 可能适合摘要章节
- 提交信息冗余 —— 部分团队偏好描述性提交信息
False Positive Warnings
误报警告
Do NOT flag these as AI-generated:
| Pattern | Why It's Valid |
|---|---|
| "Ensure" in security docs | Standard term for security requirements |
| "Comprehensive" in test coverage discussion | Accurate technical descriptor |
| Formal tone in API reference docs | Expected register for reference material |
| "Leverage" in financial/business domain code | Domain-specific meaning, not AI filler |
| Bold formatting in CLI help text | Standard convention |
| Structured intro paragraphs in RFCs/ADRs | Expected format for these document types |
"This module provides" in Python | Idiomatic Python module docstring |
| Rhetorical questions in blog posts | Appropriate for informal content |
以下模式请勿标记为AI生成内容：
| 模式 | 有效原因 |
|---|---|
| 安全文档中的"Ensure" | 安全要求的标准术语 |
| 测试覆盖率讨论中的"Comprehensive" | 准确的技术描述符 |
| API参考文档中的正式语气 | 参考资料的预期语体 |
| 金融/业务领域代码中的"Leverage" | 领域特定含义，非AI冗余词汇 |
| CLI帮助文本中的粗体格式 | 标准惯例 |
| RFC/ADR中的结构化引言段落 | 此类文档的预期格式 |
Python | Python模块文档字符串的惯用写法 |
| 博客文章中的修辞疑问句 | 适合非正式内容 |
Integration
集成
With beagle-core:review-verification-protocol
beagle-core:review-verification-protocol与beagle-core:review-verification-protocol
集成
beagle-core:review-verification-protocolBefore reporting any finding:
- Read the surrounding context (full paragraph or function)
- Confirm the pattern is AI-characteristic, not just formal writing
- Check if the project has established conventions that match the pattern
- Verify the suggestion improves clarity without changing meaning
在报告任何发现前：
- 阅读周围上下文（完整段落或函数）
- 确认该模式是AI文本的典型特征，而非仅仅是正式写作
- 检查项目是否有与该模式匹配的既定惯例
- 验证建议在不改变语义的前提下提升了清晰度
With beagle-core:llm-artifacts-detection
beagle-core:llm-artifacts-detection与beagle-core:llm-artifacts-detection
集成
beagle-core:llm-artifacts-detectionCode-level patterns (tautological docstrings, obvious comments) overlap with 's style criteria. When both skills are loaded:
llm-artifacts-detection- focuses on writing style (how it reads)
review-ai-writing - focuses on code artifacts (whether it should exist at all)
llm-artifacts-detection - If exists, skip findings already captured there
.beagle/llm-artifacts-review.json
代码层面的模式（同义反复的文档字符串、冗余注释）与的风格标准存在重叠。当同时加载两个技能时：
llm-artifacts-detection- 聚焦于写作风格（文本的可读性）
review-ai-writing - 聚焦于代码制品（内容是否必要）
llm-artifacts-detection - 若存在，则跳过已在其中记录的发现
.beagle/llm-artifacts-review.json
Output Format
输出格式
Report each finding as:
text
[FILE:LINE] ISSUE_TITLE
- Category: content | vocabulary | formatting | communication | filler | code_docs
- Type: specific_pattern_name
- Original: "the problematic text"
- Suggestion: "the improved text" or "delete"
- Risk: Low | Medium
- Fix Safety: Safe | Needs review每个发现按以下格式报告：
text
[FILE:LINE] ISSUE_TITLE
- Category: content | vocabulary | formatting | communication | filler | code_docs
- Type: specific_pattern_name
- Original: \"问题文本\"
- Suggestion: \"改进后的文本\" 或 \"delete\"
- Risk: Low | Medium
- Fix Safety: Safe | Needs reviewRisk Levels
风险等级
- Low — Filler phrases, obvious comments, emoji. Removing improves clarity with no meaning change.
- Medium — Vocabulary swaps, structural changes, docstring rewrites. Meaning could shift if done carelessly.
- 低 —— 冗余短语、明显的注释、表情符号。删除可提升清晰度且不改变语义。
- 中 —— 词汇替换、结构调整、文档字符串重写。若处理不当可能改变语义。
Fix Safety
修复安全性
- Safe — Mechanical replacement or deletion. No judgment needed.
- Needs review — Rewrite requires understanding context. Human should verify the replacement preserves intent.
- 安全 —— 机械替换或删除。无需判断。
- 需审查 —— 重写需要理解上下文。需人工验证替换是否保留了原意。