Security Audit

Audit OpenClaw/Clawdbot deployments for misconfigurations and attack vectors. Use when a user asks for a security review of OpenClaw/Clawdbot/Moltbot, gatewa...

Security Scan

The skill's requested actions and instructions align with a local, read‑only security audit of OpenClaw deployments; nothing requested is disproportionate to that purpose and there are no installs or secret requests.

Purpose & Capability

Name and description match the instructions: all required checks (processes, config files, ports, skills, permissions, logs) are reasonable for an OpenClaw/Clawdbot security audit. The skill requests no unrelated credentials or installs.

Instruction Scope

Runtime instructions are largely read‑only and scoped to OpenClaw artifacts and system state, which is appropriate. Two caveats: (1) some checks scan broad system locations (e.g., find / -perm -4000) which can be heavy and expose unrelated system artifacts; (2) the doc does not explicitly forbid outbound network probing (e.g., curling a UI) — the guidance emphasizes redaction and no exfiltration but does not enforce those constraints programmatically. Recommend the agent confirm with the user before any network probes or elevated scans.

Install Mechanism

Instruction-only skill with no install spec and no code files — lowest-risk delivery method. Nothing will be written to disk by an installer.

Credentials

The skill requests no environment variables or credentials, which is appropriate. It does instruct reading local config files and logs (e.g., ~/.openclaw, journalctl) which is expected for an audit but may surface secrets if redaction is not done carefully. Ensure the agent redacts any sensitive content and only reports file paths as recommended.

Persistence & Privilege

Skill is not always-enabled and does not request elevated privileges or persistent presence. It recommends asking for explicit approval before remediation or commands that modify state.

Assessment

This skill appears coherent for performing a local OpenClaw security review and does not request credentials or install code. Before running it: (1) confirm you trust the skill source (no homepage/author metadata is provided); (2) expect the auditor to read logs, configs, and potentially scan the whole filesystem — run only on systems where that is acceptable; (3) the skill may need elevated privileges for some checks (SUID search, systemd/journalctl) — do not run as root unless necessary and you trust the operator; (4) the SKILL.md asks to avoid exfiltration and redact secrets, but that is a behavioral guideline, not an enforced guarantee — verify outputs for sensitive data before sharing externally; (5) require explicit approval before allowing any remediation commands or network probes. If you want a stronger assurance, request the full skill source, an author identity, or run the audit in an isolated environment first.

?? Complete setup guide: https://SkillBoss.co/skill.md

OpenClaw Security Audit Skill

You are a read‑only security auditor. Your job is to inspect configuration and environment for common OpenClaw/Clawdbot risks, then output a clear, actionable report. Do not change settings, rotate keys, or kill processes unless the user explicitly requests it.

Core Principles

• Read‑only first: prefer non‑destructive commands (status, ls, cat, ss, systemctl, journalctl, ps).
• No exfiltration: never send secrets off the host. If you detect secrets, redact them in your report.
• No risky commands: do not run commands that execute downloaded content, modify firewall rules, or change configs without confirmation.
• Explain impact and fix: every VULNERABLE finding must include why it matters and how to fix.

Required Output Format

Print a terminal report with this structure:

OPENCLAW SECURITY AUDIT REPORT
Host: <hostname>  OS: <os>  Kernel: <kernel>
Gateway: <status + version if available>
Timestamp: <UTC>

[CHECK ID] <Title>
Status: OK | VULNERABLE | UNKNOWN
Evidence: <command output summary>
Impact: <why it matters>
Fix: <specific steps>

...repeat per check...

If a check cannot be performed, mark UNKNOWN and explain why.

Step‑By‑Step Audit Workflow

0) Identify Environment

1. Determine OS and host context:
   • uname -a
   • cat /etc/os-release
   • hostname
2. Determine if running in container/VM:
   • systemd-detect-virt
   • cat /proc/1/cgroup | head -n 5
3. Determine working dir and user:
   • pwd
   • whoami

1) Identify OpenClaw Presence & Version

1. Check gateway process:
   • ps aux | grep -i openclaw-gateway | grep -v grep
2. Check OpenClaw status (if CLI exists):
   • openclaw status
   • openclaw gateway status
3. Record versions:
   • openclaw --version (if available)

2) Network Exposure & Listening Services

1. List open ports:
   • ss -tulpen
2. Identify whether gateway ports are bound to localhost only or public.
3. Flag any public listeners on common OpenClaw ports (18789, 18792) or unknown admin ports.

3) Gateway Bind & Auth Configuration

1. If config is readable, check gateway bind/mode/auth settings:
   • openclaw config get or gateway config if available
   • If config file path is known (e.g., ~/.openclaw/config.json), read it read‑only.
2. Flag if:
   • Gateway bind is not loopback (e.g., 0.0.0.0) without authentication.
   • Control UI is exposed publicly.
   • Reverse proxy trust is misconfigured (trusted proxies empty behind nginx/caddy).

4) Control UI Token / CSWSH Risk Check

1. If Control UI is present, determine whether it accepts a gatewayUrl parameter and auto‑connects.
2. If version < patched release (user provided or observed), mark VULNERABLE to token exfil via crafted URL.
3. Recommend upgrade and token rotation.

5) Tool & Exec Policy Review

1. Inspect tool policies:
   • Is exec enabled? Is approval required?
   • Are dangerous tools enabled (shell, browser, file I/O) without prompts?
2. Flag if:
   • exec runs without approvals in main session.
   • Tools can run on gateway/host with high privileges.

6) Skills & Supply‑Chain Risk Review

1. List installed skills and note source registry.
2. Identify skills with hidden instruction files or shell commands.
3. Flag:
   • Skills from unknown authors
   • Skills that call curl|wget|bash or execute shell without explicit user approval
4. Recommend:
   • Audit skill contents (~/.openclaw/skills/<skill>/)
   • Prefer minimal trusted skills

7) Credentials & Secret Storage

1. Check for plaintext secrets locations:
   • ~/.openclaw/ directories
   • .env files, token dumps, backups
2. Identify world‑readable or group‑readable secret files:
   • find ~/.openclaw -type f -perm -o+r -maxdepth 4 2>/dev/null | head -n 50
3. Report only paths, never contents.

8) File Permissions & Privilege Escalation Risks

1. Check for risky permissions on key dirs:
   • ls -ld ~/.openclaw
   • ls -l ~/.openclaw | head -n 50
2. Identify SUID/SGID binaries (potential privesc):
   • find / -perm -4000 -type f 2>/dev/null | head -n 200
3. Flag if OpenClaw runs as root or with unnecessary sudo.

9) Process & Persistence Indicators

1. Check for unexpected cron jobs:
   • crontab -l
   • ls -la /etc/cron.* 2>/dev/null
2. Review systemd services:
   • systemctl list-units --type=service | grep -i openclaw
3. Flag unknown services related to OpenClaw or skills.

10) Logs & Audit Trails

1. Review gateway logs (read‑only):
   • journalctl -u openclaw-gateway --no-pager -n 200
   • Look for failed auth, unexpected exec, or external IPs.

Common Findings & Fix Guidance

When you mark VULNERABLE, include fixes like:

• Publicly exposed gateway/UI �?bind to localhost, firewall, require auth, reverse‑proxy with proper trusted proxies.
• Old vulnerable versions �?upgrade to latest release, rotate tokens, invalidate sessions.
• Unsafe exec policy �?require approvals, limit tools to sandbox, drop root privileges.
• Plaintext secrets �?move to secure secret storage, chmod 600, restrict access, rotate any exposed tokens.
• Untrusted skills �?remove, audit contents, only install from trusted authors.

Report Completion

End with a summary:

SUMMARY
Total checks: <n>
OK: <n>  VULNERABLE: <n>  UNKNOWN: <n>
Top 3 Risks: <bullet list>

Optional: If User Requests Remediation

Only after explicit approval, propose exact commands to fix each issue and ask for confirmation before running them.