OpenClaw Security Hardening - Security Suite¶

Publisher: reason-machines/security-skills | Installs: 874 | Category: Security

Automated security hardening for OpenClaw agents. Audits file permissions, network exposure, environment variables, and skill configurations - then applies security best practices. Ships with companion openclaw-security-watchdog (761 installs) for continuous monitoring.

Prerequisites¶

OpenClaw installed and running
macOS or Linux (file permission hardening is OS-specific)
Node.js 18+
npx available

Installation¶

Install both the hardening tool and watchdog together:

npx skills add reason-machines/security-skills/openclaw-security-hardening
npx skills add reason-machines/security-skills/openclaw-security-watchdog

Configuration¶

{
  "security": {
    "hardening": {
      "filePermissions": true,
      "envScanning": true,
      "networkRestriction": true,
      "skillAudit": true,
      "strictMode": false
    },
    "watchdog": {
      "enabled": true,
      "scanInterval": "1h",
      "alertOnViolation": true,
      "autoFix": false
    }
  }
}

Setting	Default	Description
`filePermissions`	`true`	Lock down config files to owner-only
`envScanning`	`true`	Scan environment for hardcoded secrets
`networkRestriction`	`true`	Audit outbound network access
`skillAudit`	`true`	Review installed skills for known vulnerabilities
`strictMode`	`false`	Enforce all findings (may break some integrations)
`watchdog.autoFix`	`false`	Auto-apply fixes (set `true` for production)

Capabilities¶

File Permission Hardening¶

Scans ~/.openclaw/, skill directories, and config files
Sets owner-only permissions (600/700) on sensitive files
Detects world-readable tokens, keys, and secrets
Flags config files with group/other read access

Environment Scanning¶

Checks for hardcoded API keys in environment variables
Flags passwords and tokens in shell config files
Detects .env files with world-readable permissions
Recommends migration to secret managers

Network Restriction¶

Audits which domains the agent connects to
Flags connections to untrusted/unexpected hosts
Recommends allowlisting for production agents
Detects plaintext (non-TLS) connections

Skill Audit¶

Scans installed skills for known vulnerability patterns
Checks skill permissions and access scopes
Flags skills with excessive filesystem/network access

CLI Reference¶

Command                           Description
──────────────────────────────────────────────────
security-hardening audit          Run full security audit
security-hardening apply          Apply hardening recommendations
security-hardening permissions    Audit file permissions only
security-hardening env-scan       Scan for hardcoded secrets
security-hardening network        Audit network access

security-watchdog start           Start continuous monitoring
security-watchdog status          Show current security posture
security-watchdog alerts          View recent alerts

CorpusIQ Use Cases¶

Production agent deployment: Run hardening audit before deploying any agent to production
Compliance preparation: Generate security posture reports for SOC 2 / ISO 27001 audits
Multi-tenant environments: Ensure agent boundaries between customer data partitions
Incident response: Watchdog alerts on unexpected file modifications or network connections

Troubleshooting¶

Audit found too many issues¶

# Start without strict mode
security-hardening audit --strict=false

# Apply only critical findings first
security-hardening apply --severity=critical

# Review each finding interactively
security-hardening audit --interactive

Permission hardening broke a skill¶

# Check which file was locked down
security-hardening permissions --verbose

# Restore permissions on specific path
chmod 755 ~/.openclaw/skills/<skill-name>

# Add exception to config

Watchdog false positives¶

# Add allowed domains
security-watchdog allowlist --add api.trusted-service.com

# Tune scan sensitivity
security-watchdog config --sensitivity=medium

← Security Hardening Setup Guide | Discovery Page →

↑ Skills Catalog

Part of the Hermes Skills Library - curated by CorpusIQ. Content remains attributed to original authors and repositories. CorpusIQ - one MCP endpoint, all your business tools.

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects