I built ClawShield after noticing that many OpenClaw deployments expose agents publicly without authentication or container hardening.

ClawShield performs static security checks on:

Public bind addresses

Disabled authentication

Containers running as root or privileged

API key exposure in .env files

World-writable config files

It produces deterministic, schema-versioned JSON output and supports CI enforcement via severity thresholds (--fail-on).

The architecture is intentionally simple:

Scanners → Facts → Policy Engine → Findings

No remote calls, no telemetry, no mutation of configs.

The goal isn’t to be a black-box security scanner, but a transparent policy verification layer for agent runtimes.

Would appreciate feedback on:

Additional high-signal checks

Policy design

Runtime security for AI agents generally

Happy to answer any questions about design decisions