Security Hardener

Audit and harden OpenClaw configuration for security. Scans openclaw.json for vulnerabilities, exposed credentials, insecure gateway settings, overly permiss...

Security Scan

The skill appears to do exactly what it claims — a local, read-only security auditor and hardener for OpenClaw configs that scans openclaw.json, checks permissions, and can apply fixes — with no network exfiltration or unrelated credential requests.

Purpose & Capability

Name/description match the implementation: the tool inspects OpenClaw config files, checks gateway/auth/exec/agent settings, scans for API-key patterns, and checks file permissions. No unrelated binaries, installs, or external services are required.

Instruction Scope

SKILL.md and the script restrict operations to local config files (default ~/.openclaw/openclaw.json or a supplied path) and config directory permissions. The 'fix' command (per README/SKILL.md) will modify local config files (it claims to create backups first) — this is expected for a hardener but users should review fixes before applying them.

Install Mechanism

No install spec or external downloads — the skill is instruction-only with a bundled Python script. This minimizes supply-chain risk.

Credentials

No environment variables or credentials are requested. The script scans for many common API-key formats (Anthropic, OpenAI, Google, GitHub, Slack, etc.), which is appropriate for a secret scanner targeting OpenClaw configs.

Persistence & Privilege

always is false; the skill does not request persistent/platform-wide privileges. It operates locally and only modifies files when the user runs 'fix'. Autonomous invocation is permitted by default for skills but is not combined with other red flags here.

Assessment

This tool appears coherent and local-only, but be cautious before running automatic fixes: 1) Run an initial 'audit' (read-only) and/or 'audit -f json' to review findings. 2) Inspect the generated report and any suggested fixes; consider backing up your config manually even if the script claims to create backups. 3) If you run 'fix', review the script or the backup to confirm changes are safe. 4) Note the secret scanner may produce false positives; verify any 'exposed keys' before rotating credentials. 5) Because the tool suggests moving keys to ~/.openclaw/.env, ensure that file is created and restricted (chmod 600). If you want extra assurance, run the script in a controlled environment or inspect scripts/hardener.py in full before invoking write operations.

Audit your OpenClaw configuration and apply security best practices automatically.

Quick Start

# Full security audit (read-only, no changes)
python scripts/hardener.py audit

# Audit a specific config file
python scripts/hardener.py audit --config /path/to/openclaw.json

# Audit with JSON output
python scripts/hardener.py audit -f json

# Auto-fix issues (creates backup first)
python scripts/hardener.py fix

# Fix specific issues only
python scripts/hardener.py fix --only gateway,permissions

# Scan for exposed credentials in config
python scripts/hardener.py scan-secrets

# Generate a security report
python scripts/hardener.py report -o security-report.md

# Check file permissions
python scripts/hardener.py check-perms

Commands

Command	Args	Description
audit	[--config PATH] [-f FORMAT]	Full security audit (read-only)
fix	[--config PATH] [--only CHECKS]	Auto-fix issues (with backup)
scan-secrets	[--config PATH]	Scan for exposed API keys/tokens
report	[-o FILE]	Generate detailed security report
check-perms	[--config-dir PATH]	Check file permissions

Security Checks

Check	Severity	Description
gateway-bind	CRITICAL	Gateway not bound to loopback
exposed-keys	CRITICAL	API keys in config instead of .env
insecure-auth	HIGH	allowInsecureAuth or dangerouslyDisableDeviceAuth enabled
exec-sandbox	HIGH	exec sandbox mode not set to restricted
file-perms	HIGH	Config files readable by others (not 600)
agent-allow-all	MEDIUM	agentToAgent.allow: ["*"] is overly permissive
no-heartbeat	MEDIUM	No heartbeat configured (can't detect outages)
no-session-reset	MEDIUM	No session reset policy (memory leak risk)
no-pruning	LOW	No context pruning (cost and performance impact)
no-memory-flush	LOW	Memory flush disabled (context loss on pruning)

Scoring

The audit produces a security score from 0-100:

• 90-100: Excellent — production-ready
• 70-89: Good — minor improvements recommended
• 50-69: Fair — several issues to address
• 0-49: Poor — critical issues require immediate attention

Example Output

╔══════════════════════════════════════════════════╗
║  OPENCLAW SECURITY AUDIT                         ║
╠══════════════════════════════════════════════════╣
║  Score: 75/100 (Good)                            ║
║                                                  ║
║  ✅ Gateway bound to loopback                    ║
║  ✅ No exposed API keys in config                ║
║  ⚠️  exec sandbox mode: unrestricted             ║
║  ⚠️  agentToAgent allow: * (too permissive)      ║
║  ❌ File permissions too open (644 → should be 600) ║
║  ✅ Heartbeat configured                         ║
║  ✅ Session reset policy active                   ║
║  ⚠️  No context pruning configured               ║
╚══════════════════════════════════════════════════╝