smartbrainactivity/smartbrain-skill-auditor ? reverse-engineered prompt

Build me a small Node.js command line tool that scans a local folder of AI Skills before I run them, and warns me if anything looks suspicious.

I want it to check common risky patterns in JavaScript, Python, shell, and Markdown, like dynamic code execution, hidden network calls, destructive commands, base64 or other obfuscation, hardcoded secrets, and sneaky prompt injection text inside markdown files. It should also give me a clear summary at the end, list each file it flagged, and explain why it thinks something is risky.

Please keep it dependency free, easy to run from the terminal, and let the rules live in a separate config file so I can tweak them later without rewriting the scanner. It should scan folders recursively, work on local files only, and be simple enough that I can point it at any downloaded skill directory. If you need to check current Node docs or best practices online, go ahead and do that.