init-harness

name	init-harness
description	Creates AI agent task management structure with feature backlog (ai/tasks/), TDD enforcement, and progress tracking. Use when setting up agent-foreman, initializing feature-driven development, creating task backlog, or enabling TDD mode. Triggers on 'init harness', 'setup feature tracking', 'create feature backlog', 'enable strict TDD', 'initialize agent-foreman'.
allowed-tools	["Bash","Read","Glob","Grep","Write","Edit"]
user-invocable	true

⚡ Init Harness

One command: agent-foreman init

Quick Start

agent-foreman init

Creates: ai/tasks/, ai/progress.log, ai/init.sh, CLAUDE.md

TDD Mode (Default: Recommended)

During init, you'll be prompted for TDD mode. Recommended is the default (tests suggested but not required).

User Says	TDD Mode	Effect
"strict TDD" / "require tests"	strict	Tests REQUIRED - check/done fail without tests
"recommended" / "optional tests" / (default)	recommended	Tests suggested but not enforced
"disable TDD" / "no TDD"	disabled	No TDD guidance

The prompt auto-skips after 10 seconds with recommended mode.

Modes

Mode	Command	Effect
Merge (default)	agent-foreman init	Keep existing + add new features
Fresh	agent-foreman init --mode new	Replace all features
Preview	agent-foreman init --mode scan	Show without changes

Task Types

Type	Command	Use Case
Code (default)	agent-foreman init	Software development
Ops	agent-foreman init --task-type ops	Operational tasks, runbooks
Data	agent-foreman init --task-type data	ETL, data pipelines
Infra	agent-foreman init --task-type infra	Infrastructure provisioning
Manual	agent-foreman init --task-type manual	Manual-only verification

Auto-Detection

1. ARCHITECTURE.md exists → use it (fast)
2. Source code exists → AI scan + auto-save ARCHITECTURE.md
3. Empty project → generate from goal

Pre-Init (Recommended)

For existing projects:

agent-foreman init --analyze    # First: understand project
agent-foreman init              # Then: create harness

Created Files

ai/
├── tasks/              # Task backlog (modular markdown)
│   ├── index.json      # Task index
│   └── {module}/       # Module directories
│       └── {id}.md     # Individual tasks
├── progress.log        # Session audit log
├── init.sh             # Bootstrap script
└── capabilities.json   # Detected test/lint/build
CLAUDE.md               # AI agent instructions
docs/ARCHITECTURE.md    # Auto-generated architecture doc