Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
A Python library for structured task execution in AI agents. Plans are defined as ordered steps, validated for safety issues, optionally approved by a human, then executed with support for dry runs, retries, and rollback checkpoints. Requires only the Python standard library.
Validation, approval gates, dry-run mode, and rollback checkpoints ship in a single file with zero dependencies, so it drops into any Python agent without conflict.
When to use it
- Gating database migrations behind a human approval step
- Dry-running a destructive cleanup script before committing to it
- Orchestrating multi-API workflows with per-step retry logic
- Flagging irreversible operations before an automated pipeline runs
- Saving and reloading task plans across agent sessions
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Run an ETL pipeline that extracts records from a database, transforms them, and loads them into a warehouse
- 1Define three steps (extract, transform, load) with expected outputs and rollback flags set
- 2Call validate_plan to surface dangerous operations and collect warnings
- 3Execute the plan with dry_run=True to simulate without side effects
- 4Approve the plan and execute it with stop_on_error=True
- 5Call get_execution_summary to report progress percentage and per-step results
Execution summary showing 100% progress and success status for each step