技能详情（站内镜像，无评论）
许可证：MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本：v0.1.0
统计：⭐ 0 · 602 · 1 current installs · 1 all-time installs
⭐ 0
安装量（当前） 1
🛡 VirusTotal ：可疑 · OpenClaw ：可疑
Package：actualcwhitlock/humanizer-2
安全扫描（ClawHub）
- VirusTotal ：可疑
- OpenClaw ：可疑
OpenClaw 评估
The skill's code and tooling match its stated purpose, but the SKILL.md/docs include explicit instructions to override model/system prompts (an 'always-on' behavior) and the package contains runnable servers — together these are out-of-band behaviors you should review before installing or enabling globally.
目的
The name/description (humanize AI-generated text) aligns with the included code: analyzers, humanizer logic, CLI, API server, and MCP server. However the registry metadata lists this as instruction-only while the bundle contains many runnable code files (api server, mcp-server). That mismatch (no install spec despite runnable servers) is noteworthy: the codebase expects local installation/run which the metadata doesn't declare.
说明范围
SKILL.md and related docs do exactly what the skill claims (pattern detection and rewriting), but they also include explicit 'Always-On Mode' guidance that tells users to add rules directly into system prompts / custom instructions (e.g., 'NEVER use these words'). Those lines amount to system-prompt override instructions which change an LLM's global behavior beyond per-invocation use. The pre-scan also flagged 'system-prompt-override' in SKILL…
安装机制
No install spec is declared in the registry (instruction-only), but the package includes a README, package.json, and multiple runnable components (api-server, mcp-server) that assume 'npm install' and 'node' executions. This is not high-risk by itself, but it's an inconsistency the user should be aware of: installing/run steps are manual and the code will create network-accessible servers if you run them.
证书
The skill declares no required environment variables, no credentials, and no config paths. The code reviewed does not demand secrets. That is proportionate to the stated functionality (text analysis/humanization).
持久
always:false (good), but documentation explicitly recommends adding the tool's rules to system prompts or custom instructions ('Always-On Mode'), and the repo provides code (MCP/API servers) that can be integrated into other LLM clients. If you follow the doc's 'Always-On' advice or wire the MCP/API into your LLM environment, the skill effectively gains persistent influence over model outputs. This combination (documentation instructing system…
安装（复制给龙虾 AI）
将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「Humanizer」。简介：Humanize AI-generated text by detecting and removing patterns typical of LLM ou…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/actualcwhitlock/humanizer-2/SKILL.md
（来源：yingzhi8.cn 技能库）SKILL.md
暂无本地缓存内容，可在后台执行详情同步。