harness-work

Automate end-to-end Harness v3 task execution from Plans.md workflows.

2|Updated Mar 30, 2026
One-click install
npx skills add https://github.com/fockus/claude-skill-build --skill harness-work-fockus
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: harness-work
Source: https://github.com/fockus/claude-skill-build/tree/main/skills/harness-work
Command: npx skills add https://github.com/fockus/claude-skill-build --skill harness-work-fockus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Harness-work integrates multiple legacy execution skills into a single unified workflow, handling planning, implementation, breezing, and parallel team runs for Harness v3 tasks to reduce manual coordination and errors.

Core Features & Use Cases

  • Unified execution: orchestrates Plans.md tasks from a single task to full parallel team execution, including solo, parallel, and breezing modes.
  • TDD-friendly workflow: supports test-first development, automated reviews, and rich status reporting to keep work moving smoothly.
  • Flexible orchestration: adapts to user intent with flags like --parallel, --breezing, and --codex, and auto-mode logic when no explicit mode is provided.

Quick Start

Run harness-work to automatically execute Plans.md tasks end-to-end, or override mode with --parallel, --breezing, or --codex.

Frequently Asked Questions about harness-work

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

FAQPage Schema
How do I automate end-to-end task execution for Plans.md workflows?

Automating Plans.md workflows is handled by unifying planning, implementation, review, and release into a single execution path. It processes tasks end-to-end using Solo, Parallel, or Breezing modes with explicit mode flags.

What is the best way to run parallel team execution on Harness v3 tasks?

Running parallel team execution on Harness v3 tasks is achieved using the --parallel mode flag. This orchestrates multiple Plans.md tasks concurrently while enforcing TDD and automated code reviews.

How do I enforce TDD and automated code reviews in a continuous execution pipeline?

Enforcing TDD and automated reviews in an execution pipeline is supported natively by applying test-first development and codex/reviewer-based checks before git-based commits and Plans.md updates.

Does this unified execution workflow support automated CI recovery actions?

Automated CI recovery actions are supported directly within the workflow. The execution generates rich completion reports and applies recovery logic automatically when continuous integration failures occur.

Can I use auto-mode logic for Plans.md tasks without specifying explicit execution flags?

Auto-mode logic applies when no explicit mode is provided, automatically determining the appropriate execution path for your Plans.md tasks. You can override this automatic selection using --parallel, --breezing, or --codex flags.