/init

name: init
description: Initialize task harness with ai/tasks/ directory structure
allowed-tools: [Bash, Read, Glob, Grep, Write, Edit, Skill]
argument-hint: "[--mode new|scan] [--task-type ops|data|infra|manual]"

STEP 0: INVOKE SKILL (MANDATORY)

**BEFORE doing anything else, you MUST invoke the `init-harness` skill:**

Skill({ skill: "init-harness" })

The instructions below are a fallback only if the skill fails to load.

---

FALLBACK: EXECUTE NOW

Run this command immediately:

agent-foreman init

Wait for completion. Do not interrupt.

**TDD Mode Prompt**: During init, you will be asked about TDD mode. **Default: recommended mode** (tests suggested but not required). The prompt auto-skips after 10 seconds with default.

TDD Mode Configuration

| User Says | TDD Mode | Effect | |-----------|----------|--------| | "strict TDD" / "require tests" / "enforce TDD" | `strict` | Tests REQUIRED for all features | | "recommended" / "optional tests" / "no strict" / (default) | `recommended` | Tests suggested but not enforced | | "disable TDD" / "no TDD" | `disabled` | No TDD guidance |

When prompted "Select TDD enforcement level":

• Press **1** for strict mode - tests required
• Press **2** (default) for recommended mode - tests optional
• Press **3** for disabled - no TDD guidance
• Wait 10s for auto-skip with recommended mode

Context-Based Behavior

The command auto-detects and handles:

| Context | Behavior | |---------|----------| | `docs/ARCHITECTURE.md` exists | Use it for fast init | | Source code exists | AI scan + auto-save ARCHITECTURE.md | | Empty project | Generate features from goal | | `ai/tasks/` exists | Merge mode (keep existing + add new) |

If User Specifies Mode

| User Says | Execute | |-----------|---------| | "fresh" / "new" / "replace" | `agent-foreman init --mode new` | | "preview" / "scan" / "dry-run" | `agent-foreman init --mode scan` | | (default) | `agent-foreman init` |

Task Type Option

For non-code projects, specify the task type:

| User Says | Execute | |-----------|---------| | "ops" / "operational" / "runbook" | `agent-foreman init --task-type ops` | | "data" / "ETL" / "pipeline" | `agent-foreman init --task-type data` | | "infra" / "infrastructure" | `agent-foreman init --task-type infra` | | "manual" / "checklist" | `agent-foreman init --task-type manual` | | (default) | `agent-foreman init` (code type) |

After Completion

Report what was created:

• `ai/tasks/` - Task backlog (modular markdown)
• `ai/progress.log` - Session log
• `ai/init.sh` - Bootstrap script
• `CLAUDE.md` - AI instructions