soul-guardian

**Judgement:** A well-scoped, security-focused integrity guard for critical OpenClaw workspace files. It’s a strong fit when you care about protecting governance/identity files (SOUL.md, AGENTS.md, etc.) from silent drift, with minimal workflow changes. **What it does well** - Detects and reports drift against approved baselines using SHA-256 hashes. - Auto-restores SOUL.md and AGENTS.md to their last approved state and alerts the user, reducing exposure to malicious or accidental edits. - Provides chat-ready alert output ideal for HEARTBEAT integration. - Maintains a tamper-evident audit log with hash chaining and unified diffs, enabling post-incident review and explicit approval of intentional changes. - Uses safer implementation choices (no symlinks, atomic writes) to reduce certain filesystem risks. **Key risks / limitations** - **Threat model limits:** If an attacker controls both the workspace and the state directory, this system cannot reliably protect integrity or logs. - **Attribution limits:** It cannot prove who made a change; actor metadata is only best-effort. - **Config friction:** Without a clear process for approving legitimate edits, users may experience friction (restores of intended changes) or may over-approve, weakening protection. **Recommended scenarios** - Long-running or multi-agent OpenClaw setups where SOUL.md / AGENTS.md define core behavior and must not drift silently. - Higher-stakes workflows where users want immediate, visible alerts in TUI/chat when key files change. - Environments that value auditability: being able to review diffs and an integrity-protected log of file changes over time. If you need a focused integrity layer for a small set of critical files and are willing to maintain baselines and approvals, this skill is a strong, low-overhead choice.