using-superharness

Enforce TDD, dual review, and git worktree isolation for AI coding tasks.

4|Updated Apr 5, 2026
One-click install
npx skills add https://github.com/Mrlyk/superharness --skill using-superharness
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: using-superharness
Source: https://github.com/Mrlyk/superharness/tree/main/skills/using-superharness
Command: npx skills add https://github.com/Mrlyk/superharness --skill using-superharness

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Superharness prevents ad-hoc, unsafe, or unverified AI-driven code changes by mechanically enforcing a disciplined software engineering workflow at session start and throughout task execution. It closes gaps where agents might skip tests, drift from requirements, or declare work done without verification.

Core Features & Use Cases

  • Session-start injection: Automatically injects project conventions and the dispatch protocol into every AI session so the agent follows predefined rules before doing work.
  • Process enforcement: Enforces TDD, requirement clarification, two-stage review, QA validation, task isolation via git worktrees, and trace logging to maintain auditability.
  • Convention discovery & task tracking: Scans the codebase to populate .superharness/spec/ and tracks task state and traces in .superharness/tasks/ for reproducible development.
  • Use Case: Start a new feature development session and have the agent generate a vetted plan, implement via failing tests first, run dual reviews, and produce verifiable merge-ready commits.

Quick Start

Invoke the using-superharness skill at session start to inject project conventions, enable TDD enforcement, run skill checks, and orchestrate the full development workflow.

Frequently Asked Questions about using-superharness

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

FAQPage Schema
How do I enforce TDD and code review workflows in AI-driven coding sessions?

Enforcing TDD and dual code review in AI coding sessions requires injecting project conventions at session start, mandating failing tests before implementation, and blocking completion until QA validation passes. This workflow prevents unverified or ad-hoc AI code changes.

What is the best way to isolate AI coding tasks using git worktrees?

Isolating AI coding tasks with git worktrees involves creating separate working directories for each task, tracking state in dedicated task directories, and logging traces to maintain auditability. This ensures reproducible development without branch conflicts.

How do I automate requirement clarification before an AI agent starts coding?

Automating requirement clarification involves triggering a dispatch protocol at session start that forces the AI agent to brainstorm, generate a vetted plan, and discover project conventions before writing any code. This prevents requirement drift during task orchestration.

Can I track AI development task state and traces for reproducible workflows?

Tracking AI development task state and traces is achieved by scanning the codebase to populate specification directories and recording task progress in dedicated tracking folders. This produces auditable artifacts for reproducible development workflows.

Why does my AI agent skip tests and declare work done without verification?

AI agents skip tests and declare unverified work done when lacking mechanical process enforcement. Injecting a strict dispatch protocol at session start closes these gaps by forcing test-first development, two-stage review, and QA validation before completion.

Does using-superharness require external dependencies to enforce development workflows?

Using-superharness requires no external dependencies to enforce development workflows. It operates by scanning the codebase to discover conventions and generating specification and task tracking artifacts to maintain process-level requirements mechanically.