AI 文本净化器(中文版)

v1.0.0

中文AI文本检测与改写工具。当用户需要检测AI生成文本、优化AI文本使其更自然、降低AI痕迹、文本去重、论文降重时使用。支持检测16+类AI特征,自动改写冗余表达,清理Markdown格式,移除chatbot痕迹,输出详细报告和AI概率分数。

0· 114·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name/description match what the package implements: detection and rewriting of Chinese text. Required artifacts (rules, synonyms, detector and rewriter modules, CLI scripts) are appropriate and present. No unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md instructs running local Python scripts (detect.py, transform.py, compare.py) with file input/output and optional user rules. The code implements exactly those operations (reading files, applying rules, writing output). Instructions do not direct the agent to read unrelated system files, send data to external endpoints, or access credentials.
Install Mechanism
No install spec. This is instruction+code that runs locally; all code is included and uses only local file I/O and standard library. Optional spaCy integration is present but only used if the library and model are installed—no network downloads or remote installs are invoked by the scripts themselves.
Credentials
No required environment variables, no credentials, and no config paths requested. The tool accepts user-supplied JSON rule files and writes output to user-specified paths (expected for this functionality).
Persistence & Privilege
Skill is not forced-always, does not request persistent privileges, and does not modify other skills or system-wide agent settings. It only reads input files, optional user rule files, and writes output files provided by the user.
Assessment
This package appears to do what it claims: a rule-based Chinese AI-text detector and rewriter that runs locally. Before using: (1) Back up original files — aggressive mode and replacements can remove or alter content, possibly affecting nuance or factual details. (2) Expect non-deterministic synonym/rewrites due to random choices; run repeatedly if you need deterministic output. (3) The scripts read any file path you pass and write outputs to paths you specify — avoid pointing them at sensitive system files. (4) Optional spaCy usage only activates if spaCy and a Chinese model are installed; the package itself does not download or exfiltrate data. If you need additional assurance, review the included rule JSONs (resources/zh_rules.json) and synonyms to confirm replacements you want/ do not want, or run the detector in JSON/score-only mode first.

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

latestvk97ewtxcx911jdd7a8fcmcs7t583kz4j

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments