clawaudit 0.2.0a6

OpenClaw security audit tool

Foresight ClawAudit

Security audit tool for OpenClaw instances. Scans runtime configuration, system environment, Skill plugins, supply chain integrity, and known CVEs — covering 31 check points across 8 categories.

What OpenClaw Leaves on Your System

Type	Path	Content
Main config	~/.openclaw/	openclaw.json, API keys, workspace, memory
Workspace	~/.openclaw/workspace/	Files the AI operated on, Skill plugin code
Memory	~/.openclaw/memory/	SQLite + Markdown memory data
Agent auth	~/.openclaw/agents/*/agent/auth-profiles.json	OAuth tokens, third-party authorizations
Logs	~/.openclaw/logs/	Gateway runtime logs
Legacy dirs	~/.clawdbot/, ~/.moltbot/, ~/.molthub/	Pre-rename configs (Clawdbot -> Moltbot -> OpenClaw)
System service	LaunchAgent (macOS) / systemd (Linux)	Gateway daemon on 127.0.0.1:18789, runs 24/7
Exec approvals	~/.openclaw/exec-approvals.json	Shell command allowlist

Foresight ClawAudit helps you identify security risks across all of these.

Quick Start

# Install uv if not already installed (https://docs.astral.sh/uv/)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Run directly without install Foresight ClawAudit
uvx clawaudit scan --mode local

Platform Requirements

• Linux / macOS: Supported natively
• Windows: Must run under WSL. Native Windows is not supported

Installation

pip install clawaudit
# or
uv pip install clawaudit

Usage

# Local scan (auto-detect OpenClaw path)
clawaudit scan --mode local
clawaudit scan --mode local --path ~/.openclaw

# Filter by category or check
clawaudit scan --mode local --category NET,AUTH

# Output: terminal (default) / json / sarif / markdown
clawaudit scan --mode local --format json -o report.json

# CVE database
clawaudit cvedb update
clawaudit cvedb list

Root is not required. Checks needing elevated privileges will SKIP with a reason. Exit codes: 0 = all pass, 1 = has FAIL, 2 = has CRITICAL.

Check Categories

Code	Category	Checks	Summary
NET	Network Exposure	4	Port binding, TLS, IOC
AUTH	Authentication	2	Weak token, UI bypass
CONF	Configuration	4	Permissions, flags, baseline
EXEC	Execution Sandbox	8	Injection, sandbox, prompt
SC	Supply Chain	3	Source, integrity, poisoning
DATA	Data Security	4	Keys, leakage, exfiltration
SYS	System Security	6	Privileges, cron, SSH
CVE	Known Vulnerabilities	Dynamic	Version matching (271 advisories)

License

Proprietary. For personal and internal use only. Redistribution and modification are not permitted.

About

Developed by Beijing Institute of AI Safety and Governance (Beijing-AISI). Beijing-AISI focuses on safe development, safety and security assessment, and governance of AI.

Contact

• Email: [e-mail]
• Website: https://beijing-aisi.ac.cn