humanizer-zh
v1.0.0去除文本中的 AI 生成痕迹。适用于编辑或审阅文本，使其听起来更自然、更像人类书写。 基于维基百科的"AI 写作特征"综合指南。检测并修复以下模式：夸大的象征意义、 宣传性语言、以 -ing 结尾的肤浅分析、模糊的归因、破折号过度使用、三段式法则、 AI 词汇、否定式排比、过多的连接性短语。
⭐ 18· 8.5k·196 current·206 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (去除 AI 写作痕迹) match the SKILL.md and README: both describe detection and rewriting patterns for Chinese text. No unrelated binaries, env vars, or credentials are requested.
Instruction Scope
SKILL.md gives detailed, prescriptive rewriting rules and examples — all within the stated purpose. It also mentions processing file contents (e.g., '请人性化 article.md 文件中的内容') and allows Read/Write/Edit tools; this is reasonable for a text-editing skill but means the agent may access local files you direct it to. The guide encourages injecting personality and sometimes first‑person phrasing, which can intentionally alter tone and factual framing — not a security problem but a behavior you should expect and verify on sensitive material.
Install Mechanism
No install spec or third‑party downloads are included; the README suggests normal git/npx clone instructions for the user. Nothing in the package writes or fetches arbitrary code at runtime.
Credentials
The skill requests no environment variables, credentials, or config paths. The required tool privileges (Read/Write/Edit) are proportional to a text-rewriting/editing skill.
Persistence & Privilege
always:false and default invocation policies are used. The skill does not request persistent system-wide privileges or modify other skills. Autonomous invocation is allowed by platform default but is not combined with other concerning access.
Assessment
This skill appears coherent and low-risk: it only contains instructions for spotting and rewriting AI-like phrasing and requests no credentials or network access. Before using: (1) test it on non-sensitive text to confirm rewrites preserve facts you care about, because the guide explicitly encourages changing tone and sometimes factual framing; (2) be careful when asking it to process local files — only point it at files you want rewritten; (3) confirm where any edited output is saved and avoid feeding highly confidential data to third‑party skill installs. If you need guarantees that wording won't change factual claims, ask the skill to only adjust style (not content) or request a diff/summary of edits first.Like a lobster shell, security has layers — review code before you run it.
latestvk979ftchea2g6301xbm61xy9qx80w3vs
License
MIT-0
Free to use, modify, and redistribute. No attribution required.