Security Audit

Checks installed OpenClaw version against known CVEs (gateway auth bypass, 1-click RCE, WebSocket token leak).

Fetches the live OpenClaw CVE feed and checks installed version against known vulnerabilities with remediation guidance.

Detects running NanoClaw instances via processes, containers, and channel connections.

Detects running OpenClaw instances via Docker containers, processes, and port 18789 binding.

Audits ~/.config/nanoclaw/mount-allowlist.json for permissions, blocked paths, and security settings.

Audits ~/.openclaw directory and openclaw.json for permissions, gateway binding, auth, discovery, and credential exposure.

Audits NanoClaw container security: non-root user, capabilities, resource limits, read-only rootfs.

Advanced configuration checks: sandbox mode, exec host policy, DM scope/policy, filesystem restriction, log redaction.

Checks .env file permissions in common OpenClaw installation locations.

Verifies NanoClaw credential isolation: env var exposure, CLAUDE.md secrets, session credential safety.

Scans config files, shell history, MEMORY.md, and session logs for exposed credentials. Also detects prompt injection in MEMORY.md (poisoned instructions, base64, invisible Unicode, C2 callbacks) and session transcripts (jailbreak phrases, encoded payloads).

Checks for network egress restrictions on NanoClaw containers. Warns about unrestricted outbound access (known vulnerability).

Verifies NanoClaw channel authentication: WhatsApp, Telegram, Slack, Discord credential security.

Scans skills directories for dangerous install commands, secret injection, data exfiltration patterns, prompt-level attack instructions, companion hook file analysis, file integrity drift detection, and openclaw.json skills config audit.

Scans session JSONL files for suspicious bash commands executed by AI agents — detects data exfiltration (curl POST, wget), reverse shells, base64-to-shell execution, privilege escalation (chmod 777, setuid), sensitive file access (/etc/shadow, authorized_keys), download-and-execute patterns, environment variable dumping, and history clearing.

Checks SOUL.md permissions, sensitive data, prompt injection patterns, base64, unicode, and file size.

Audits running OpenClaw Docker container for user, capabilities, privileged mode, network, resource limits, and volume mounts.

Checks elevated tool access, browser control, group access policy, plugin allowlist, and trusted proxy configuration.

Checks permissions on credential directories, OAuth profiles, session transcript stores, and log files. Content-scans log files for leaked credentials and exploitation patterns.

Detects NemoClaw installation via Python package, config directory, and running processes.

Audits NeMo Guardrails config.yml for input rails, output rails, jailbreak detection, and PII masking.

Verifies NemoClaw audit logging is enabled for compliance and security monitoring.

Verifies NemoClaw role-based access control, tool permissions, and privilege separation.

Verifies NemoClaw PII detection and masking configuration for compliance (GDPR, HIPAA, SOC2).