NullClaw | Autonomous AI Ecosystem

Security

Security controls are implemented in src/security/* and enforced during runtime/tool execution.

Core Security Modules

  • pairing.zig: pairing and token guards
  • policy.zig: command/path risk controls
  • sandbox.zig + adapters (landlock, firejail, bubblewrap, docker, detect)
  • audit.zig and tracker.zig: security-relevant event tracking
  • secrets.zig: secret handling helpers

Default Security Posture

  • local-first bind defaults
  • pairing required by default
  • policy checks before sensitive operations
  • path and command restrictions on tool execution

Operational Verification

nullclaw doctor
nullclaw status
nullclaw capabilities --json

High-Impact Config Keys

  • gateway.require_pairing
  • gateway.allow_public_bind
  • autonomy.allowed_commands
  • autonomy.allowed_paths
  • security.sandbox.backend
  • security.audit.*

Production Guidance

  • keep pairing enabled outside trusted localhost workflows
  • avoid broad command/path allowlists
  • treat capabilities output as final authority for what can execute