soul-guardian

What This Skill Does

Soul Guardian serves as an automated integrity watchdog for your OpenClaw agent workspace. By maintaining cryptographic baselines for critical configuration files, it ensures that your agent's identity, behavior, and memory structures remain consistent. The skill monitors protected files for unauthorized modifications and provides two distinct defensive postures: automated restoration for critical system files like SOUL.md and AGENTS.md, or reactive alerting for operational files. It includes a built-in audit trail and audit verification, ensuring that any drift—whether accidental or malicious—is caught, logged, and surfaced for immediate user review.

Installation

To integrate Soul Guardian, execute the following command in your terminal: clawhub install openclaw/skills/skills/davida-ps/soul-guardian Once installed, initialize your workspace baseline by running python3 skills/soul-guardian/scripts/soul_guardian.py init --actor setup --note "initial baseline". It is highly recommended to add the heartbeat check snippet to your HEARTBEAT.md to ensure your agent proactively reports its integrity status during its normal operational cycle.

Use Cases

• Agent Governance: Prevent accidental modification of core identity files that govern agent behavior.
• Security Hardening: Detect and revert unauthorized workspace changes that might attempt to jailbreak or alter the agent's core instructions.
• Operational Stability: Ensure that configuration files like TOOLS.md or MEMORY.md haven't been corrupted during agent execution, maintaining a reliable environment for complex multi-step workflows.

Example Prompts

1. "Soul Guardian, run a status check on all core files and report if any drift has been detected since the last baseline."
2. "I am planning to update my SOUL.md file. Can you help me finalize the changes and then re-run the Soul Guardian init command to set the new baseline?"
3. "Soul Guardian, display the latest audit log entries and verify the integrity of the current workspace hashes."

Tips & Limitations

• Proactive Maintenance: Always run approve immediately after making intentional configuration changes; otherwise, the auto-restore feature will revert your work.
• Granularity: Use the ignore functionality for high-churn directories like memory/ to prevent alert fatigue, while keeping a tight lock on system-critical YAML or Markdown files.
• Environment: Ensure the skill has the necessary filesystem permissions to read and write to your workspace directory to allow for automatic restoration of tampered files.