Humanizer

What to consider before installing or enabling this skill: - Treat the code as executable: although the registry lists this as instruction-only, the package contains runnable servers (api-server, mcp-server). Only run the code after reviewing package.json and all src files locally. - Do NOT blindly paste the 'Always-On' or 'NEVER use these words' sections into any agent/system prompts or your global custom-instructions. Those lines are a form of system-prompt override — they persistently change model behavior and could have unintended side effects across conversations. - If you want to use the tool, prefer on-demand invocation (run the CLI or call the local API only when needed) rather than applying global system-prompt modifications. - Review networking/exposure: api-server binds to a port and sets Access-Control-Allow-Origin: '*' (CORS open). If you deploy it publicly, verify authentication and CORS restrictions to avoid exposing text to unintended callers. - Audit dependencies and omitted files: inspect package.json and run npm audit/scan, and skim all src/*.js for outbound network calls, telemetry, or unexpected file I/O before running. - Run in isolation first: execute the tool in a sandbox or isolated environment and run the provided tests (npm test) to confirm behavior matches description. If you want a higher-confidence verdict, provide the package.json at repo root, and the full contents of src/analyzer.js, src/humanizer.js, and any other src files that were truncated — that lets a reviewer check for network calls, hidden endpoints, or credential use. If you do need persistent model behavior, implement it at the integration layer under controlled conditions rather than pasting the skill's 'Always-On' text into your system prompt.