article-taster

v1.0.0

文章品鉴师 - 多维度评估文章质量、检测AI味/大便味、识别原创内容

0· 126·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for forealmy/tmp70s6amg4.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "article-taster" (forealmy/tmp70s6amg4) from ClawHub.
Skill page: https://clawhub.ai/forealmy/tmp70s6amg4
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install tmp70s6amg4

ClawHub CLI

Package manager switcher

npx clawhub@latest install tmp70s6amg4
Security Scan
VirusTotalVirusTotal
Pending
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name/description (article quality / AI‑detection) match the provided modules: classifier, analyzers, AI detector, scorer, and report generator. No unrelated credentials, binaries, or config paths are requested. The included requirements (jieba, scikit-learn, numpy) are plausible for text analysis in Chinese and align with the skill's purpose.
Instruction Scope
SKILL.md and main.py limit runtime actions to analyzing text provided via --text or files the user explicitly supplies. The code reads input text, performs heuristic analysis, and outputs JSON/Markdown; there are no instructions to read arbitrary system configuration, harvest environment variables, or send results to external endpoints. The only file I/O is reading user-provided article files (via --file or --dir) and normal local operations for generating reports.
Install Mechanism
There is no install spec in the registry (instruction-only), but the bundle includes Python source and a requirements.txt. Running the skill as intended will require installing Python dependencies; this is expected but worth noting. No remote downloads, URL-based installers, or extract operations are present in the package metadata.
Credentials
The skill declares no required environment variables or credentials and the code does not reference external secrets. No cross-service tokens or unrelated credentials are requested. The only external requirement is typical Python packages listed in requirements.txt.
Persistence & Privilege
Flags show always:false and normal agent invocation. The skill does not request permanent system presence nor attempts to modify other skills or global agent configuration. It does not request elevated privileges.
Assessment
This skill appears internally consistent with its description. Before installing or running: 1) Review and, if possible, run the code in an isolated environment (virtualenv or container) and install dependencies from requirements.txt. 2) Do not pass sensitive local files (password files, SSH keys, system configs) to the --file or --dir options — the tool will read any file path you give it. 3) If you plan to deploy it in production or allow autonomous agent invocation, monitor outbound network activity (there is no network code in the provided files, but it’s good practice). 4) If you need stricter guarantees, request provenance (who authored/published the skill) or run a security audit on the repository code (lint, dependency checks).

Like a lobster shell, security has layers — review code before you run it.

latestvk974ncc49f31tbx2gfr58gny0584hhq4
126downloads
0stars
1versions
Updated 2w ago
v1.0.0
MIT-0

Article Taster - 文章品鉴师

帮助用户提前品尝文章可读性,过滤低质量内容,节省宝贵阅读时间

核心定位

"文章含大质量检测" —— 站在文章分析师的角度,多维度评估文章价值:

  • 技术文章:衡量技术含量、学习价值
  • AI生成检测:正确识别原创内容(古诗等不被误判)
  • 情感散文:分析情感曲线、架构模式
  • 小说:分析情节结构(但不剧透)

工作流程

输入文章文本/标题
       ↓
┌─────────────────────────────────────┐
│  M1: 文章类型识别                     │
│  技术文章 | 情感散文 | 小说 | 其他     │
└─────────────────────────────────────┘
       ↓
┌─────────────────────────────────────┐
│  M2: 专业分析 (根据类型分发)           │
│  ├─ 技术文章 → TechAnalyzer          │
│  ├─ 情感/散文 → CreativeAnalyzer     │
│  ├─ 小说 → NovelAnalyzer (无剧透)   │
│  └─ 其他 → GenericAnalyzer           │
└─────────────────────────────────────┘
       ↓
┌─────────────────────────────────────┐
│  M3: AI生成检测与原创识别              │
│  ├─ 文本统计特征                      │
│  ├─ 困惑度分析                        │
│  ├─ 风格一致性                        │
│  └─ 原创内容豁免 (古诗词/经典文学)     │
└─────────────────────────────────────┘
       ↓
┌─────────────────────────────────────┐
│  M4: 综合评分与阅读建议                │
│  最终评分 + 等级 + 阅读价值建议        │
└─────────────────────────────────────┘
       ↓
输出: JSON报告 + Markdown摘要

评分体系

技术文章评分维度

维度权重说明
技术深度30%原理剖析、实战经验、解决方案复杂度
结构清晰度25%逻辑组织、层次分明
实用性20%可操作、可落地
原创性15%独特观点、非AI生成
可读性10%表达流畅、图文配合

情感散文/小说评分维度

维度权重说明
情感表达30%情感深度、感染力、曲线设计
文笔水平25%修辞手法、句式变化、词汇丰富度
叙事结构20%情节安排、节奏把控
创意性15%独特视角、创新表达
共鸣度10%读者情感连接

AI生成检测维度

维度权重说明
文本统计10%句子长度方差、词汇密度、标点多样性
困惑度15%语言模型困惑度(过低=疑似AI)
词汇丰富度10%Type-Token Ratio、罕见词汇比例
风格一致性10%开头词重复率、过渡词规律性
语义连贯5%指代一致性、主题集中度
特殊模式10%古诗词/经典文学豁免检测

AI味/大便味检测维度 (新增)

维度权重说明
段落一致性15%AI文章段落长度高度一致,像模具铸出来的
废话率10%每句都对但空洞,无实质信息密度
模板化程度10%三段式、综上所述、首先其次最后等
人类标记5%踩坑、血泪史、说实话等personal voice

AI味评分等级

评分等级说明
<20人类写作几乎无AI味
20-40轻度疑似AI可能有轻微AI辅助
40-60中度疑似AI明显AI特征
60-80高度疑似AI强烈AI味
>80大便味极强AI味,内容空洞

综合评分公式

最终评分 = 加权得分 × 类型匹配度 × (1 - AI概率 × 0.3)

评分等级

等级分数阅读建议
A+90-100极力推荐!值得反复研读
A80-89强烈推荐!内容扎实
B+70-79推荐阅读,有价值
B60-69可读,碎片时间可看
C40-59一般,可选择性阅读
D<40不推荐,浪费时间

AI生成检测豁免规则

为防止误判真正的原创内容,采用豁免机制:

类型检测特征豁免效果
古诗词五言/七言、押韵、意象密度AI阈值提高50%+
经典文学长句、复合句、修辞丰富方差阈值放宽100%

古诗测试预期: "床前明月光..." → 识别为 classical_poetry,得分>85,标注"高度可信原创"


输出报告格式

JSON 完整报告

{
  "report_id": "taster_20260408_001",
  "title": "文章标题",
  "type": "technical_article|essay|novel|other",
  "type_confidence": 0.92,
  "overall_score": 85,
  "grade": "A",
  "reading_advice": {
    "verdict": "强烈推荐!内容扎实,适合认真阅读",
    "target_audience": "初中级开发者",
    "time_estimation": "10分钟",
    "key_benefits": ["深入浅出的架构设计", "实操性强"],
    "suitable_moments": ["专注阅读", "深度学习"]
  },
  "dimension_scores": {
    "technical_depth": {"score": 88, "weight": 0.30},
    "structure": {"score": 85, "weight": 0.25},
    "practicality": {"score": 82, "weight": 0.20},
    "originality": {"score": 78, "weight": 0.15},
    "readability": {"score": 80, "weight": 0.10}
  },
  "ai_detection": {
    "ai_probability": 0.15,
    "is_ai_generated": false,
    "ai_flavor_score": 38,
    "ai_flavor_level": "轻度疑似AI",
    "confidence_label": "高度可信原创",
    "dimensions": {
      "text_statistics": {"score": 75},
      "perplexity": {"score": 80},
      "vocabulary_richness": {"score": 72},
      "style_consistency": {"score": 68},
      "semantic_coherence": {"score": 85},
      "special_patterns": {"score": 95, "exemption_type": "classical_poetry"},
      "paragraph_uniformity": {"score": 70},
      "bullshit_ratio": {"score": 85},
      "template_patterns": {"score": 90},
      "human_markers": {"score": 80}
    },
    "ai_flavor_warnings": [
      "段落长度高度一致,AI味特征明显",
      "缺少人类真实表达痕迹"
    ]
  },
  "detailed_analysis": {
    "spoiler_warnings": [],
    "genre_specific": {...}
  },
  "timestamp": "2026-04-08T10:00:00Z"
}

Markdown 用户报告

# 文章品鉴报告

## 基本信息
- **标题**: 技术架构设计原则
- **类型**: 技术文章 (置信度: 92%)
- **评分**: 85分 (A级)

## 综合评价
强烈推荐!内容扎实,适合认真阅读

**目标读者**: 初中级开发者
**预计阅读时间**: 10分钟

## 维度评分
| 维度 | 得分 | 权重 |
|------|------|------|
| 技术深度 | 88 | 30% |
| 结构清晰度 | 85 | 25% |
| 实用性 | 82 | 20% |
| 原创性 | 78 | 15% |
| 可读性 | 80 | 10% |

## AI检测结果
- **AI生成概率**: 15%
- **AI味评分**: 38 (轻度疑似AI)
- **结论**: 高度可信原创

### AI味警告
- 段落长度高度一致,AI味特征明显
- 缺少人类真实表达痕迹

## 阅读建议
- 适合专注阅读、深度学习
- 核心价值: 深入浅出的架构设计、实操性强

Skill 结构

article-taster/
├── SKILL.md                    # 本文件
├── scripts/
│   ├── __init__.py
│   ├── main.py                 # 主入口
│   ├── article_classifier.py   # M1: 类型识别
│   ├── tech_analyzer.py        # M2-T: 技术文章分析
│   ├── creative_analyzer.py    # M2-C: 情感/散文分析
│   ├── novel_analyzer.py       # M2-N: 小说分析 (无剧透)
│   ├── ai_detector.py          # M3: AI生成检测
│   ├── scorer.py               # M4: 综合评分
│   └── report_generator.py     # 报告生成
├── config/
│   ├── scoring_weights.json    # 评分权重
│   ├── type_keywords.json      # 类型关键词
│   ├── ai_patterns.json        # AI检测模式
│   └── exemption_rules.json    # 原创豁免规则
├── references/
│   ├── scoring_methodology.md  # 评分方法论
│   └── spoiler_free_principles.md # 无剧透分析原则
├── requirements.txt
└── README.md

使用方式

# 分析单篇文章
python -m article_taster analyze --text "文章内容..."
python -m article_taster analyze --file article.txt

# 批量分析
python -m article_taster batch --dir ./articles

# 仅获取快速评分
python -m article_taster quick --text "文章内容..."

# 指定文章类型
python -m article_taster analyze --text "..." --type technical_article

依赖项

  • Python 3.10+
  • jieba (中文分词)
  • scikit-learn (文本相似度)
  • openai / anthropic (可选,LLM辅助评分)

Comments

Loading comments...