Kimchi Generate

<command_purpose> Create a task breakdown from requirements and research, with dependencies, complexity estimates, and traceability to requirements. </command_purpose>

Input

Read .kimchi/CONTEXT.md, .kimchi/REQUIREMENTS.md, and .kimchi/RESEARCH.md. If any are missing, tell the user which prerequisite to run.

Process

1. Map Requirements to Tasks

For each v1 requirement, identify the implementation tasks needed:

• •Database changes (migrations, schema)
• •Service/business logic
• •API endpoints / controllers
• •UI components / views
• •Tests
• •Configuration

2. Identify Dependencies

Build a dependency graph:

• •Which tasks must complete before others can start?
• •Which tasks can run in parallel?
• •Are there any circular dependencies? (Error if so)

3. Estimate Complexity

If any task is L, consider splitting it into smaller tasks.

4. Write PLAN.md

Write to .kimchi/PLAN.md:

# Plan: [Feature Name]

**Generated:** [today's date]
**Source:** .kimchi/CONTEXT.md, .kimchi/REQUIREMENTS.md, .kimchi/RESEARCH.md

## Task Overview

| ID | Task | Complexity | Depends On | Requirements |
|----|------|------------|------------|--------------|
| 001 | [Task title] | S | - | [REQ-ID] |
| 002 | [Task title] | M | - | [REQ-ID] |
| 003 | [Task title] | M | 002 | [REQ-ID] |

## Dependency Graph

[ASCII dependency graph showing parallel and sequential tasks]

## Task Details

### Task 001: [Title]
**Complexity:** S
**Requirements:** [REQ-ID]
**Depends On:** None

[Description of what this task does]

**Deliverables:**
- [What files/changes will be produced]

**Context Needed:**
- [What existing code the agent needs to see]

---

### Task 002: [Title]
**Complexity:** M
**Requirements:** [REQ-ID]
**Depends On:** None

[Description]

**Deliverables:**
- [List]

**Context Needed:**
- [List]

---

[... additional tasks ...]

## Execution Strategy

**Parallel groups (tasks with no interdependencies):**
- Wave 1: [tasks that can start immediately]
- Wave 2: [tasks that depend on Wave 1]
- Wave 3: [tasks that depend on Wave 2]

**Integration tests:** Final task after all others complete

Report: "Plan generated. Saved to .kimchi/PLAN.md" Suggest: "Run /kimchi:review for multi-persona critique, or /kimchi:refine to skip review."

Key Principles

• •Every task traces to a v1 requirement: No orphan tasks
• •No L tasks: Split them. Large tasks fail in agent execution.
• •Dependencies are explicit: Not implied, not "obvious"
• •Context Needed lists what the agent must read: Not what's nice to know