Mindflow

Converts text, Markdown files, or Txt files into mind map images. Use this skill when users want to generate mind maps/brain maps from articles, broadcast sc...

This skill appears to be what it claims (text → markmap → HTML → image) but proceed cautiously: 1) npm install will pull markmap and Puppeteer and Puppeteer will download a Chromium binary — only install on trusted hosts and consider pinning package versions or using a lockfile. 2) The SKILL.md expects an LLM conversion step; verify where that runs (locally vs external API) because sensitive content may be sent to a model endpoint. 3) Inspect the full html-to-image.js (the provided snippet was truncated in the review) to confirm there are no unexpected network calls, remote endpoints, or shell executions before running it. 4) Run the code in a sandboxed environment (container/VM) if you must test it, and avoid feeding sensitive files until you confirm the runtime behavior. If you want, provide the complete html-to-image.js (untruncated) and any package.json or lockfile so I can re-evaluate and raise the confidence level.

Type: OpenClaw Skill Name: mindflow Version: 1.0.6 The skill exhibits risky capabilities including arbitrary local file reading and shell command execution (markmap, node) which, while aligned with its stated purpose of mind map generation, pose a security risk. Specifically, the 'scripts/html-to-image.js' utility uses Puppeteer with the '--no-sandbox' and '--disable-setuid-sandbox' flags, significantly reducing browser isolation and increasing the impact of potential XSS-to-RCE vulnerabilities if processing untrusted input. The instructions in 'SKILL.md' also direct the agent to read user-provided file paths without explicit validation or sandboxing.