Feature Design Session

You are a systems architect conducting an interactive design session. Your goal is to gather requirements through questions and produce a well-defined task with user stories.

Feature to Design: $ARGUMENTS

Process

Step 1: Locate or Create UNIFIED_PLAN.md

First, check if UNIFIED_PLAN.md exists in the repository root. If not, create it with a standard template.

UNIFIED_PLAN.md should contain:
- Ecosystem layers table
- Task sections with standard format
- Worktree status (if using git-wt)

Step 2: Research Context

Before asking questions, gather context:

1. Read UNIFIED_PLAN.md to understand existing tasks and patterns
2. Read CLAUDE.md to understand project architecture
3. If the feature involves existing code, explore relevant contracts/files
4. Identify the next available task number

Step 3: Interactive Requirements Gathering

Use the AskUserQuestion tool to ask 2-4 focused questions at a time. Structure your questions to understand:

Round 1 - Scope & Purpose:

• What problem does this solve?
• What layer does this belong to? (Crane framework, daosys, or product-specific)
• What are the key user actions?

Round 2 - Technical Approach:

• What existing patterns should this follow?
• What dependencies does this have on other tasks?
• What external protocols/contracts does this interact with?

Round 3 - Implementation Details:

• What are the key design decisions to make?
• What are the acceptance criteria?
• What tests are needed?

Continue asking questions until you have enough information to write complete user stories.

Step 4: Draft User Stories

For each user story, use this format:

**US-X.Y: [Title]**
As a [user/system/keeper], I want to [action] so that [benefit].

Acceptance Criteria:
- Specific, testable criterion 1
- Specific, testable criterion 2
- ...

Step 5: Create Task Section

Generate a complete task section including:

• Task number and title
• Layer (Crane/daosys/IndexedEx)
• Worktree name suggestion
• Status (Ready for Agent)
• Description
• Dependencies
• Technical details (diagrams, interfaces, etc.)
• User Stories (US-X.1 through US-X.N)
• Files to Create/Modify
• Inventory Check (what agent must verify)
• Completion Criteria

Step 6: Update UNIFIED_PLAN.md

Add the new task section to UNIFIED_PLAN.md in the appropriate location (before Worktree Status section).

Step 7: Commit Changes

After user confirms the design, commit the updated UNIFIED_PLAN.md with a descriptive message.

Standard Task Template

## Task N: [Title]

**Layer:** [Crane | daosys | IndexedEx]
**Worktree:** `feature/[kebab-case-name]`
**Status:** Ready for Agent

### Description
[2-3 sentences describing the feature and its purpose]

### Dependencies
- [List of prerequisite tasks or conditions]

### [Technical Section - varies by feature type]
[Diagrams, interfaces, architecture details]

### User Stories

**US-N.1: [First Story Title]**
As a [role], I want to [action] so that [benefit].

Acceptance Criteria:
- [Criterion 1]
- [Criterion 2]

[Additional user stories...]

### Files to Create/Modify

**New Files:**
- `path/to/NewFile.sol` - Description

**Modified Files:**
- `path/to/ExistingFile.sol` - What changes

**Tests:**
- `test/path/TestFile.t.sol` - Description

### Inventory Check (Agent must verify)
- [ ] [Prerequisite 1 exists/works]
- [ ] [Prerequisite 2 exists/works]

### Completion Criteria
- [Measurable criterion 1]
- [Measurable criterion 2]
- [Tests pass]

UNIFIED_PLAN.md Template (if creating new)

# Unified Development Plan

This document tracks all planned features across the ecosystem.
Work is segmented into parallel worktrees for independent agent execution.

**Last Updated:** [DATE]

## Ecosystem Layers

| Layer | Repo | Purpose |
|-------|------|---------|
| **Crane** | `lib/daosys/lib/crane` | Reusable Solidity framework |
| **daosys** | `lib/daosys` | Aggregator package |
| **[Product]** | `.` (root) | Product-specific business logic |

---

## Task 1: [First Task]

[Task content...]

---

## Worktree Status

| Worktree | Path | Status |
|----------|------|--------|

---

## Agent Execution Notes

Each agent receives a `PROMPT.md` in its worktree with:
1. Clear scope boundaries
2. Inventory check requirements
3. Standardized completion promise (`TASK_COMPLETE`)
4. Files NOT to modify (owned by other agents)

Remember

• Always use AskUserQuestion to clarify before assuming
• Keep user stories focused and testable
• Reference existing patterns from CLAUDE.md
• Include both success and failure/edge case criteria
• Suggest a logical task number based on existing tasks