跳到正文
TakoAPI
返回技能列表

gpt-analyzer

由 raghulpasupathi 提供

Coding Agents & IDEs
384下载量由 clawskills.sh 提供
0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

GPT-specific pattern detection with model fingerprinting and version identification.

描述

该技能的功能

Analyzes text for GPT-generated content using phrase matching, structural patterns, and sentence uniformity checks. Attempts to distinguish between GPT-4 and GPT-3.5 outputs based on model-specific phrase signatures. Returns a confidence score, detected model, and a plain-language recommendation.

Unlike generic AI detectors, it attempts model-level attribution, separating GPT-4 patterns from GPT-3.5 ones rather than returning a single binary result.

何时使用

  • Screening job applicant cover letters for AI authorship
  • Checking student essay submissions for GPT usage
  • Auditing blog or publication content before publishing
  • Detecting AI-generated product reviews or support tickets
  • Verifying whether a document draft was written by a human

示例工作流

输入

User asks: Is this essay GPT-generated?

智能体
  1. 1

    Receives the essay text as input

  2. 2

    Scans for GPT-4 and GPT-3.5 phrase signatures in the normalized text

  3. 3

    Analyzes structural patterns like numbered lists and bullet points

  4. 4

    Calculates sentence length variance to detect uniform writing style

  5. 5

    Returns confidence score, detected model version, and recommendation

输出

{ isGPT: true, confidence: 82, detectedModel: 'gpt-4', recommendation: 'Likely GPT' }

安装

clawhub install raghulpasupathi/gpt-analyzer

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