harness

Manage long-running agent workflows with state persistence and failure recovery.

7|1|Updated Jan 19, 2026
One-click install
npx skills add https://github.com/makoMakoGo/code-dispatcher-toolkit --skill harness-makomakogo
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: harness
Source: https://github.com/makoMakoGo/code-dispatcher-toolkit/tree/main/bundles/harness
Command: npx skills add https://github.com/makoMakoGo/code-dispatcher-toolkit --skill harness-makomakogo

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill provides a robust framework for long-running AI agent tasks, ensuring progress is saved and recoverable across session interruptions, context resets, and agent failures.

Core Features & Use Cases

  • Progress Checkpointing: Saves task status and intermediate results to persistent files.
  • Failure Recovery: Automatically resumes or rolls back tasks upon interruption or error.
  • Task Dependency Management: Handles complex workflows where tasks depend on each other.
  • Use Case: Running a multi-day research and analysis task where the AI needs to process large datasets, generate reports, and potentially restart without losing its place.

Quick Start

Use the harness skill to initialize a new project for durable agent work.

Frequently Asked Questions about harness

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I save agent workflow state for recovery after a session interruption?

State recovery for interrupted agent workflows is handled through automatic checkpointing, which persists task status and intermediate results to JSON files and git commits for seamless resumption.

What is the best way to manage task dependencies in long-running autonomous agent workflows?

Task dependency management in autonomous agent workflows is handled by a durable session framework that orchestrates complex task chains through both exclusive and concurrent execution modes.

Can I use Claude Code hooks for autonomous agent session management and error handling?

Yes, autonomous agent session management and error handling can be integrated with Claude Code hooks, which the framework uses to manage session lifecycle events and trigger failure rollback procedures.

Does this agent checkpointing framework support concurrent execution for parallel tasks?

Yes, the agent checkpointing framework supports concurrent execution for parallel tasks, providing both exclusive and concurrent execution modes to manage complex workflows with detailed logging.

Why do autonomous agents lose progress during context resets and how to prevent it?

Autonomous agents lose progress during context resets because intermediate states are volatile; preventing this requires durable sessions that perform atomic state updates to persistent JSON files.

When should I not use a durable agent harness for workflow state management?

A durable agent harness for workflow state management should not be used for short-lived, stateless tasks where the overhead of atomic JSON state updates and git commits outweighs failure recovery benefits.