🔐 SecureClaw

Cross-platform security scanner and remediation tool for OpenClaw / Clawdbot

SecureClaw is a comprehensive security toolkit that audits and fixes security issues in OpenClaw and Clawdbot installations. It runs 33+ checks across 18 categories, auto-fixes config-level issues, and organizes findings into severity tiers (T1-T4) so you can prioritize what matters most.

- 🔒 Credential protection (file permissions, ownership hardening)
- 🛡️ Prompt injection detection
- 🔑 Plaintext secret detection
- 👤 Privilege escalation checks
- 📦 Skill supply chain security
- 🔍 Code obfuscation detection
- 📝 Security documentation generation
- 🌐 Zero-trust tunnel detection (Tailscale, Cloudflare Tunnel)
- 🧩 OpenClaw-specific checks (commands, sandbox, plugins, agents, hooks)

Installation

``bash
npm install -g secureclaw
`

Zero npm dependencies. Requires Node >= 18.

Quick Start

$3

`bash


Auto-detect OpenClaw/Clawdbot installation

secureclaw audit

Scan specific path

secureclaw audit /path/to/openclaw

Output as JSON, SARIF, or HTML

secureclaw audit --format json --output report.json
secureclaw audit --format sarif --output report.sarif
secureclaw audit --format html --output report.html

Filter by minimum severity

secureclaw audit --severity high
`

$3

`bash


Interactive mode (approve each fix)

secureclaw fix

Automatic mode (apply all fixes)

secureclaw fix --auto

Fix only critical-tier issues

secureclaw fix --auto --tier T1

Fix critical + high-risk issues

secureclaw fix --auto --tier T2
`

$3

`bash


Quick overview without full scan output


secureclaw summary
`

Features

$3

1. Control Panel Security - Gateway authentication, token strength, IP binding, zero-trust tunnel detection
2. Prompt Injection Protection - Detects missing prompt-guard skill
3. Plaintext Secrets - Finds API keys, tokens in config files
4. Skill Supply Chain - Scans for malicious patterns, exfiltration domains
5. Privilege Configuration - Checks for root/admin execution
6. Security Documentation - Validates SECURITY.md presence
7. Code Obfuscation - Detects base64, eval(), exec() usage
8. Credential Encryption - Validates encryption at rest
9. Browser Control Exposure - RCE risk, JavaScript evaluation
10. Filesystem Permissions - World-readable configs
11. DM Pairing - Open DM policy, session isolation
12. Group Mention Gating - Group chat attack surface
13. Model Provider Privacy - Data retention warnings
14. Tool Policy Enforcement - Unrestricted tools, elevated permissions
15. Incident Response - Audit logging, log redaction
16. MCP Server Security - Tool poisoning detection
17. Advanced Features - Shell commands, config editing, debug mode, webhooks, sandbox, plugins, agent-to-agent messaging
18. OpenClaw-Specific - Tailscale Funnel, insecure auth, context segregation

$3

Findings are organized into four tiers for prioritized remediation:

| Tier | Name | Description | Examples |
|------|------|-------------|----------|
| T1 | Critical Exposure | Immediate remote exploitation risk | No gateway auth, shell commands in chat, public Tailscale Funnel |
| T2 | High Risk | Significant security gaps | No context segregation, debug commands enabled, no plugin allowlist |
| T3 | Hardening | Defense-in-depth improvements | Sandbox read-write access, network-enabled containers, browser JS eval |
| T4 | Best Practice | Operational maturity | Log redaction patterns, session isolation, elevated tool restrictions |

$3

- Generate strong gateway tokens (64-char cryptographic)
- Generate webhook authentication tokens
- Disable dangerous chat commands (bash, config, debug, restart)
- Disable agent-to-agent messaging
- Set sandbox workspace to read-only
- Isolate sandbox network (set to none)
- Disable shell environment sourcing
- Disable Tailscale Funnel (switch to serve mode)
- Configure Tailscale zero-trust networking
- Setup Cloudflare Tunnel (cloudflared)
- Install prompt-guard skill
- Enable context segregation
- Generate security documentation (SECURITY.md)
- Secure credentials (chmod 600 + ownership hardening)
- Enable audit logging

$3

| Format | Flag | Use Case |
|--------|------|----------|
| Text | --format text | Terminal output (default) |
| JSON | --format json | Programmatic consumption, CI/CD |
| SARIF | --format sarif | GitHub Code Scanning integration |
| HTML | --format html | Shareable reports with inline CSS |

$3

- ✅ Linux (Debian, Ubuntu, RHEL, CentOS)
- ✅ macOS (Homebrew)
- ✅ Windows (WSL)

$3

17 out of 20 auto-fix actions work without root/sudo. When a fix fails due to permissions, SecureClaw shows a boxed copy-paste command:

`
[ERROR] Permission denied: Install Tailscale

┌─ Manual fix (copy & paste) ──────────────────────
│ curl -fsSL https://tailscale.com/install.sh | sh && sudo tailscale up
└─────────────────────────────────────────────────
`

Commands are platform-aware (Linux gets apt/curl, macOS gets brew). Config edits use node -e one-liners since Node.js is guaranteed present.

Usage Examples

$3

`bash
$ secureclaw audit ~/.openclaw
`

Output:
`
╔═══════════════════════════════════════════════════════════════╗
║ SecureClaw - Clawdbot Security Suite v1.5.0 ║
╚═══════════════════════════════════════════════════════════════╝

[INFO] Running security audit...
[INFO] Target: /home/user/.openclaw

SecureClaw Security Scanner v1.5.0
Platform: linux

[1/18] Checking Control panel security...
[2/18] Checking Prompt injection protection...
...
[17/18] Checking Advanced features...
[18/18] Checking OpenClaw-specific security...

═══════════════════════════════════════════════════════
SECURITY ASSESSMENT COMPLETE
═══════════════════════════════════════════════════════

Findings:
Critical: 5
High: 14
Medium: 9
Low: 3

Risk Score: 141 (SEVERE)
`

$3

`bash


Fix only T1 (critical exposure) issues first

$ secureclaw fix --auto --tier T1

Tier filter: T1 (showing T1)
Fixable issues: 7/31

[SUCCESS] Gateway token updated (xjQMc1rL...)
[SUCCESS] Set gateway.tailscale.mode = "serve"
[SUCCESS] Webhook token generated (RUQQ4Y49...)
[SUCCESS] Set commands.bash = false

Fixed: 4
Failed: 3 (require sudo/system access)
`

$3

`bash
$ secureclaw fix
`

Prompts for approval on each fix:
`
═══════════════════════════════════════════════════════
Finding 1/31
═══════════════════════════════════════════════════════
Severity: CRITICAL
Tier: T1
Title: Shell commands enabled in chat via ! prefix

[FIX] Setting commands.bash = false...
Apply this fix? [Y/n]: y

[SUCCESS] Set commands.bash = false
`

$3

`bash
$ secureclaw audit --format sarif --output results.sarif


Upload to GitHub via: gh api repos/OWNER/REPO/code-scanning/sarifs


`

Risk Scoring

SecureClaw assigns risk scores based on finding severity:

- CRITICAL: 10 points each
- HIGH: 5 points each
- MEDIUM: 2 points each
- LOW: 1 point each

Risk Levels:
- 0-10: MINIMAL
- 11-20: LOW
- 21-30: MODERATE
- 31-50: HIGH
- 51+: SEVERE

Configuration

SecureClaw auto-detects OpenClaw and Clawdbot installations. When you provide a target path, it searches the target first, then falls back to the home directory:

`
Search order (first match wins):
1. /.openclaw/openclaw.json ← Docker/container installs
2. /openclaw.json
3. /.clawdbot/clawdbot.json
4. /config.json
5. /clawdbot.json
6. ~/.openclaw/openclaw.json ← Home directory fallback
7. ~/.clawdbot/clawdbot.json
`

When no target path is given, auto-detection checks:
- ~/.openclaw
- ~/.clawdbot
- /opt/clawdbot
- Common Node.js global module paths

Override with explicit path:
`bash
secureclaw audit /custom/path
`

$3

Create .secureclawrc.json in your project root to set defaults:

`json
{
"format": "json",
"severity": "medium",
"tier": "T2",
"ignore": ["no-security-md", "no-audit-log"],
"output": "secureclaw-report.json"
}
`

Exit Codes

- 0: Success, no critical issues
- 1: Error during execution
- 2: Critical security issues found (audit mode)

Requirements

- Node.js: v18.0.0 or higher
- Permissions: Some fixes require sudo/admin access
- OpenClaw/Clawdbot: Compatible with OpenClaw 2026.x+ and Clawdbot 2024.x+

Security

Found a security issue in SecureClaw itself? Report it to: security@buguard.io

License

MIT License - see LICENSE file for details

Support

📧 Email: youssef@buguard.io

Changelog

See CHANGELOG.md for the full version history.

$3

- Config detection searches target path first, home directory as fallback
- Scanner checks /.openclaw/openclaw.json` (Docker/container layouts)
- Filesystem permissions and vault checks use detected config path
- 88 tests

$3

- Platform-aware copy-paste manual fix commands in error output
- Removed non-existent dependencies (clawdbot-keyutils, universal-vault.sh)
- All external references verified against live registries
- Credential protection via file permissions instead of fake vault tools
- Non-root support: 17/20 fixes work without sudo

$3

See CHANGELOG.md for full history.