跳到正文
TakoAPI
返回技能列表

text-detection

由 raghulpasupathi 提供

CLI Utilities
324下载量由 clawskills.sh 提供
1星标
1浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Skills for analyzing and detecting AI-generated text content.

描述

该技能的功能

A set of tools for detecting AI-generated text, particularly content produced by GPT models. It analyzes perplexity, burstiness, sentence patterns, and structural signatures to classify text as AI-written, human-written, or mixed.

Multiple detection methods (perplexity scoring, GPT fingerprinting, ML classification) are bundled in a single configurable stack, removing the need to integrate separate services.

何时使用

  • Checking student essays for AI-generated content
  • Screening job application cover letters for ChatGPT usage
  • Flagging AI-written articles before editorial review
  • Filtering AI content in community moderation workflows
  • Verifying authenticity of user-submitted reviews or posts

示例工作流

输入

User asks: Is this 500-word essay AI-generated?

智能体
  1. 1

    Runs nlp-toolkit to calculate perplexity and burstiness scores

  2. 2

    Runs gpt-analyzer to check for GPT-specific phrase patterns and model signatures

  3. 3

    Runs pattern-matcher to flag repetitive sentence starts and uniform structure

  4. 4

    Aggregates confidence scores from all three tools

  5. 5

    Returns a verdict with per-tool scores and detected patterns

输出

{ isAI: true, confidence: 0.87, signals: ['uniform-length', 'low-burstiness', 'gpt-3.5 signature'] }

安装

clawhub install raghulpasupathi/text-detection

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接