safe-action

What This Skill Does

The safe-action skill acts as an intelligent gatekeeper for your most critical operations. Inspired by the "measure twice, cut once" philosophy, it implements a rigorous, three-stage pre-flight analysis before allowing potentially destructive commands to proceed. When triggered, the skill orchestrates a series of specialized diagnostic checks across the OpenClaw AgentUtil ecosystem: first, it pulls a domain-specific checklist to identify hidden risks; second, it queries the reversibility engine to determine if an action can be undone or if a safer alternative exists; and finally, it validates the timing against current business schedules, holidays, and system status context. By unifying these services, safe-action prevents catastrophic human error and mitigates risks inherent in high-stakes environment management, such as production deployments, large-scale database modifications, or unintended data deletion.

Installation

To integrate this essential safety layer into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/cutthemustard/safe-action

Ensure your agent has the necessary permissions to access the @agentutil MCP suite, as these are the primary interfaces for the safety checks.

Use Cases

• Database Administration: Use this before dropping columns, truncating tables, or running bulk migration scripts that could lead to data loss.
• Infrastructure Operations: Essential for production deployments, terminating cloud instances, or modifying sensitive firewall/DNS configurations.
• Access Control: Deploy this before revoking user permissions or altering security roles to prevent accidental lockouts or privilege escalation risks.
• Bulk Operations: Highly recommended for mass email sends or batch document deletions to verify intent and potential impact.

Example Prompts

1. "OpenClaw, before I delete the entire legacy-prod database, please run a safe-action check to see if this is reversible."
2. "I need to deploy these infrastructure changes to production. Can you execute a safe-action pre-flight first?"
3. "Run a safe-action analysis for changing the root directory permissions on the production server."

Tips & Limitations

• Contextual Awareness: Always provide as much detail as possible in your prompt (e.g., target environment, specific resource IDs) to improve the precision of the safety checks.
• Fallback Mechanisms: If the preferred MCP services are unreachable, the skill automatically attempts to use the HTTP fallback endpoints. Ensure your network permits traffic to *.agentutil.net.
• Human Override: This skill is an advisor, not an absolute barrier. It provides the data to make a decision, but you remain the final authority on executing high-stakes commands.