Security toolkit for OpenClaw – scanner, hardened configs, guides
Malicious OpenClaw skill scanner, but the market for hardening OpenClaw specifically is tiny.
You've been pinched. Now get unpinched. Find out if PinchTab is running on your host before someone else does.
CDP-based attack detection that EDR tools miss, backed by published research.
Security teams and developers running AI agent frameworks with browser access
rkhunter · chkrootkit · Lynis
Unpinched is a point-in-time scanner — think nmap for PinchTab presence. Single Go binary, no install required, runs in ~3 seconds.
It checks four things: - Local ports for a PinchTab HTTP API server (with signature verification) - Running processes matching known PinchTab binary names - Unauthenticated CDP exposure on localhost:9222 - Known filesystem artifact paths across macOS, Linux, Windows
Also ships as a GitHub Action so you can gate deploys on a clean scan result.
GitHub: https://github.com/Helixar-AI/Unpinched
The underlying research on why CDP-based attacks evade conventional security stacks is here if you're curious: https://helixar.ai/press/pinchtab-stealth-browser-attacks-yo...
Happy to answer questions on the detection logic or the threat model.
Malicious OpenClaw skill scanner, but the market for hardening OpenClaw specifically is tiny.
MCP-specific guardrails when Claude ecosystem lacks native security scanning.
Client-side scanner catches hardcoded secrets in AI code—real problem, obvious solution.
Catches malicious skills before they steal your AWS keys or pipe data exfiltration.
26 MCP-specific checks with GitHub Actions + SARIF, but confined to emerging protocol ecosystem.
NPM supply chain scanner competing against Socket, Snyk, and npm audit.