OpenClaw Security Hardening

**Judgement:** This skill is a well-designed, domain-specific security toolkit for OpenClaw that materially improves defense-in-depth around skills and workspace configuration, especially when integrated into automation (CI/cron). It is particularly strong for detecting prompt injection patterns, data exfiltration vectors, and unauthorized skill tampering via static analysis. **Key strengths:** - Multiple focused tools (scan-skills, audit-outbound, integrity-check, harden-workspace, install-guard) aligned to a clear threat model. - Non-invasive, pattern-based scanning that avoids executing untrusted skill code. - Good automation story (JSON output, exit codes, whitelist management, cron examples) and a security rules template to harden runtime behavior. **Risks / limitations:** - Pattern-based detection will inevitably have both false positives (noisy warnings) and false negatives (missed sophisticated attacks), so it should not be treated as a complete security guarantee. - The `--fix` workspace hardening may conflict with custom permission schemes or `.gitignore` strategies in complex repos and should be tested in non-production first. - Assumes a compatible shell/OS environment; behavior on non-standard or Windows environments may require adaptation. **Recommended scenarios:** - Teams running multi-skill OpenClaw installations that need basic but structured security controls against malicious or compromised skills. - Environments with moderate to high security requirements where new skills must pass automated checks before installation. - Users who want a reproducible, scriptable security baseline (including integrity monitoring and outbound domain control) rather than ad-hoc manual reviews.