Why do I need to exit Claude Code to execute the plan?

Mizu often requires different system permissions and execution environments than the Claude Code session provides, ensuring safer and more flexible autonomous operations.

How does /harness identify tasks in my plan?

It uses structured parsing to look for implementation steps and task lists, falling back to AI-assisted extraction for non-standard formats to ensure no steps are missed.

Can I run multiple plans at once?

Yes, the skill creates a unique directory within .mizu/ for every plan, keeping state, tasks, and progress isolated for each specific execution.

What is the primary purpose of the /harness skill?

It transforms a markdown plan file generated in Claude Code's plan mode into a structured execution.json file that the Mizu CLI can use to perform tasks autonomously.

Mizu Plan Harness

Name: Mizu Plan Harness
Author: brs98

bybrs98

0•

Productivity & Workflow

Transforms Claude Code plans into executable configurations for autonomous task processing with the Mizu CLI.

The Mizu Harness skill bridges the gap between planning and execution by converting Claude Code markdown plans into structured JSON configurations. It intelligently extracts tasks, infers required system permissions like Docker or AWS, and sets up verification commands to ensure code quality. By creating isolated, plan-scoped directories, it allows developers to hand off complex engineering tasks to the Mizu agent system for autonomous execution outside of the main Claude Code environment.