What is the primary purpose of the init-harness skill?

It creates a standardized task management structure that allows AI agents to track their own feature backlogs and progress using external memory files.

Can I use this on existing projects?

Yes, it includes an auto-detection mode that scans existing source code to generate architecture documentation and a corresponding feature backlog.

How does this improve Claude Code's performance?

By providing an external 'source of truth' for tasks (feature_list.json), it prevents the agent from losing track of goals during long-running or complex development sessions.

What is the difference between Strict and Recommended TDD modes?

Strict mode requires tests to be present before a task can be marked as done, while Recommended mode suggests tests without enforcing them as a hard requirement.

AI Agent Task Harness

Name: AI Agent Task Harness
Author: mylukin

bymylukin

•

217

Productivity & Workflow

Initializes a feature-driven development structure for AI agents with automated task tracking and TDD enforcement.

About

The init-harness skill sets up a robust project management framework specifically designed for AI agents using agent-foreman. It creates a dedicated 'ai/' directory containing a feature backlog, progress logs, and a bootstrap script, enabling AI agents to track their own progress across long-running tasks. By establishing clear Test-Driven Development (TDD) protocols and automatically generating architecture documentation, it ensures AI-led development remains structured, verifiable, and aligned with project goals.

Key Features

217 GitHub stars
Seamless integration with CLAUDE.md for persistent agent instructions
Automated feature backlog generation in ai/feature_list.json
Real-time session auditing and progress logging for AI interactions
Configurable TDD enforcement modes (Strict, Recommended, or Disabled)
Automatic project architecture scanning and documentation generation

Use Cases

Initializing a new AI-driven project with a clear feature roadmap and task structure
Adding external memory and feature tracking to an existing codebase for better agent autonomy
Enforcing strict TDD workflows for AI agents to ensure high code reliability

About

Key Features

217 GitHub stars
Seamless integration with CLAUDE.md for persistent agent instructions
Automated feature backlog generation in ai/feature_list.json
Real-time session auditing and progress logging for AI interactions
Configurable TDD enforcement modes (Strict, Recommended, or Disabled)
Automatic project architecture scanning and documentation generation

Use Cases

Initializing a new AI-driven project with a clear feature roadmap and task structure
Adding external memory and feature tracking to an existing codebase for better agent autonomy
Enforcing strict TDD workflows for AI agents to ensure high code reliability