Skip to content
Arxo Arxo

OpenClaw Architecture

The openclaw_architecture metric evaluates OpenClaw agent configuration (e.g. openclaw.json) for security, skill governance, observability, and supply chain risks.

OpenClaw deployments can expose risks through:

  • Config security: Gateway auth, discovery pinning, loopback binding, credentials, rate limiting
  • Skill governance: Allowlists, human approval, sandbox, tool permissions, shadow MCP servers
  • Observability: OTel plugin, audit logging, retention, GenAI semantic conventions, cost tracking, interruptibility, drift detection
  • Supply chain: ClawHub provenance, virus scan, pinned MCP versions, source integrity, egress controls

This metric runs multiple detectors and produces axis scores plus an overall health score. Run arxo metrics registry --registry-format detailed to list all detector keys and possible values; the generated Keys and possible values page is built from that schema.

  • Config Security: Gateway auth when remote, discovery pinning, loopback binding, control UI exposure, credential storage, auth token strength, DM policy, rate limiting
  • Skill Governance: Skill allowlist/denylist, self-modifying skill risk, human approval, tool poisoning, sandbox, exec security mode, agent tool over-permission, shadow MCP server
  • Observability: OTel plugin, content capture, sample rate, audit log, log retention, GenAI semantic conventions, cost tracking, interruptibility, behavioral drift detection
  • Supply Chain: ClawHub provenance, memory file write risk, metadata validation, virus scan, unpinned MCP version, skill source integrity, egress firewall, known malicious skill, typosquat names
Terminal window
arxo analyze --metric openclaw_architecture --path /path/to/openclaw-project

To dump the full detector schema (for docs or tooling):

Terminal window
arxo metrics registry --registry-format detailed --output registry.json