Skip to content

SomtoUgeh/somto-dev-toolkit / skills/prd-workflow

prd-workflow

Verified

This skill should be used when the user asks to "plan a feature", "create PRD", "interview me about requirements", "define stories", "write a spec", "break down feature into tasks", "create implementation plan", "help me scope this feature", "what questions should I answer", or discusses feature planning. Covers the PRD interview process, spec structure, story atomization, and 4-phase workflow from idea to Dex tasks.

FileKindSize
SKILL.mdskill md4.8 KB
---
name: prd-workflow
description: |
  This skill should be used when the user asks to "plan a feature", "create PRD",
  "interview me about requirements", "define stories", "write a spec", "break down
  feature into tasks", "create implementation plan", "help me scope this feature",
  "what questions should I answer", or discusses feature planning. Covers the PRD
  interview process, spec structure, story atomization, and 4-phase workflow from
  idea to Dex tasks.
version: 2.0.0
---

# PRD Workflow - Feature Planning

**Current branch:** !`git branch --show-current 2>/dev/null || echo "not in git repo"`
**Existing plans:** !`ls plans/ 2>/dev/null | head -5 || echo "none"`

The PRD (Product Requirements Document) workflow transforms ideas into atomic,
implementable stories through structured interview, research, and specification.

## When to Use PRD Workflow

- Starting a new feature from scratch
- Turning a vague idea into concrete tasks
- Need thorough requirements gathering
- Want AI-assisted research and review

## The 4-Phase Approach

| Phase | Name | Purpose |
|-------|------|---------|
| 1 | Input Classification | Identify feature name and type |
| 2 | Interview + Exploration | Gather requirements, then run research/expert agents |
| 3 | Spec Write | Create specification with Implementation Stories section |
| 4 | Dex Handoff | Parse stories into Dex tasks |

**Key Change:** Research and expert agents run after interview completes (blocking). Findings are reviewed before spec writing.

## Starting the Workflow

```bash
/prd "Add user authentication"           # From idea
/prd path/to/existing/code               # From codebase
/prd                                     # Interactive discovery
```

## Phase 2: Interview + Exploration

**Commitment:** "I will ask 8-10+ questions covering: core problem, success criteria, MVP scope, technical constraints, UX flows, edge cases, error states, and tradeoffs."

### Interview (8-10+ questions)

Cover these areas:
- Core problem, success criteria, MVP scope
- Technical systems, data models, existing patterns
- UX/UI flows, error states, edge cases
- Tradeoffs, compromises, priorities

Focus on **non-obvious** details that will bite during implementation.

### After Interview, Run Agents (blocking)

Spawn ALL agents in parallel (user waits ~2-3 min):

**Research Agents:**
- `prd-codebase-researcher` - Existing patterns, files to modify
- `git-history-analyzer` - Prior attempts, why patterns evolved
- `prd-external-researcher` - Best practices, code examples

**Expert Agents:**
- `architecture-strategist` - System design concerns
- `security-sentinel` - Auth, data exposure risks
- `spec-flow-analyzer` - User journeys, missing flows
- `pattern-recognition-specialist` - Codebase consistency

### Review Findings

Summarize key findings. Ask clarifying questions if gaps revealed.

## Phase 3: Spec Structure

Write to `plans/<feature>/spec.md` with this structure:

```markdown
# <Feature> Specification

## Overview
## Problem Statement
## Success Criteria
## User Stories
## Detailed Requirements
## Technical Design
## Edge Cases & Error Handling
## Open Questions
## Out of Scope

## Implementation Stories

### Story 1: <Title>
**Category:** functional|ui|integration|edge-case|performance
**Skills:** <skill-name>, <skill-name>
**Blocked by:** none
**Acceptance Criteria:**
- [ ] Criterion 1
- [ ] Criterion 2

## Research Findings
## Expert Review Findings
```

## Story Atomization Rules

**Commitment:** "Each story will have ≤7 acceptance criteria, touch ≤3 files, be independently testable, and have no 'and' in title."

Each story must be completable in ONE task (~15-30 min):

- **Max 7 acceptance criteria** - If more, split the story
- **Max 3 files touched** - If more, consider splitting
- **No "and" in title** - "User can X and Y" = two stories
- **Independently testable** - Can verify in isolation
- **Cleanly revertible** - Can undo without cascade

## Phase 4: Dex Handoff

Use `dex plan` to automatically create tasks from the spec:

```bash
dex plan plans/<feature>/spec.md
```

This automatically:
- Creates parent task from spec title
- Analyzes Implementation Stories section
- Generates subtasks with proper hierarchy
- Sets blocked-by relationships from "Blocked by:" lines

Verify tasks created:
```bash
dex status
dex list
```

## After PRD Completion

Use Dex for execution:

```bash
dex status              # Dashboard view
dex list --ready        # See unblocked tasks
dex start <id>          # Claim a task
/complete <id>          # Run reviewers and complete
```

## Command Reference

```bash
/prd "feature description"     # Start from idea
/prd path/to/code             # Start from codebase
/cancel-prd                   # Cancel in-progress PRD
```

## Related Commands

- `/complete` - Complete task with reviewer workflow
- `dex list` - View pending tasks
- `dex sync` - Sync with GitHub issues
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall