Promptzpromptz
LibrariesPromptsAgentsPowersSteeringHooks
Contribute
Back to Steering
steeringproduct-teams
IDECLI

Prd Guide

Nate Ober
Jan 6, 2026
Updated Jan 6, 2026
590cc3f
View on GitHub
Steering Document Content

# PRD Guide (Kiro Spec Format)

This steering file guides PRD creation with output to Kiro's native spec format. The PRD content maps to `requirements.md` and `design.md` in the spec structure.

## Input Required

Before creating PRD, ensure you have:
- Completed PRFAQ (or product concept description)
- Market research summary (if available)
- AI Framing document (for ML products)

## Dual Output Strategy

Generate TWO outputs:

1. **Traditional PRD** → `documents/PRD_[Product]_[Date].html`
2. **Kiro Spec** → `.kiro/specs/[product-slug]/requirements.md`

**Note:** The PRD document must be HTML (for browser viewing). The Kiro spec remains markdown (required by Kiro's spec format).

## EARS Syntax for Requirements

Kiro uses EARS (Easy Approach to Requirements Syntax). Convert user stories to this format:

### Ubiquitous Requirements (always true)
```
THE SYSTEM SHALL [capability]
```

### Event-Driven Requirements
```
WHEN [event occurs]
THE SYSTEM SHALL [response]
```

### State-Driven Requirements
```
WHILE [state is active]
THE SYSTEM SHALL [behavior]
```

### Conditional Requirements
```
IF [condition is true]
THE SYSTEM SHALL [action]
```

### Complex Requirements
```
WHEN [event] WHILE [state] IF [condition]
THE SYSTEM SHALL [response]
```

## Requirements.md Template

```markdown
# [Product Name] Requirements

## Overview
[2-3 sentence product description from PRFAQ]

## User Stories

### US-001: [Story Title]
**As a** [persona]
**I want to** [capability]
**So that** [benefit]

**Acceptance Criteria:**
WHEN user [action]
THE SYSTEM SHALL [response]

WHEN user [invalid action]
THE SYSTEM SHALL [error handling]

### US-002: [Story Title]
...

## Functional Requirements

### FR-001: [Requirement Title]
**Priority:** P0 | P1 | P2
**Persona:** [Primary user]

WHEN [trigger event]
THE SYSTEM SHALL [expected behavior]

**Rationale:** [Why this requirement exists]

### FR-002: [Requirement Title]
...

## Non-Functional Requirements

### NFR-001: Performance
THE SYSTEM SHALL respond to user actions within 200ms

### NFR-002: Availability
THE SYSTEM SHALL maintain 99.9% uptime

### NFR-003: Security
WHEN user session is inactive for 30 minutes
THE SYSTEM SHALL automatically log out the user

## Data Requirements

### DR-001: [Data Entity]
THE SYSTEM SHALL store [data fields]
THE SYSTEM SHALL retain data for [duration]

## Integration Requirements

### IR-001: [Integration Name]
THE SYSTEM SHALL integrate with [external system]
WHEN [trigger] THE SYSTEM SHALL [action]

## Constraints

- [Technical constraint 1]
- [Business constraint 2]
- [Regulatory constraint 3]

## Out of Scope
- [Feature explicitly excluded]
- [Capability deferred to future phase]

## Success Metrics
| Metric | Target | Measurement |
|--------|--------|-------------|
| [KPI 1] | [Value] | [How measured] |
```

## Persona to User Story Mapping

For each PRD persona, create corresponding user stories:

| Persona | Primary Workflows | User Stories |
|---------|------------------|--------------|
| [Name] | [Workflow 1] | US-001, US-002 |
| [Name] | [Workflow 2] | US-003, US-004 |

## Priority Definitions

| Priority | Meaning | Kiro Treatment |
|----------|---------|----------------|
| P0 | Must have for MVP | Include in initial spec |
| P1 | Should have | Add after P0 complete |
| P2 | Nice to have | Separate spec or backlog |

## AI/ML Product Requirements

For ML products, add these sections:

```markdown
## ML Requirements

### MLR-001: Model Performance
THE SYSTEM SHALL achieve [metric] >= [threshold]
THE SYSTEM SHALL maintain [latency] <= [value]ms

### MLR-002: Data Pipeline
WHEN new training data is available
THE SYSTEM SHALL trigger model retraining

### MLR-003: Monitoring
THE SYSTEM SHALL alert when model performance degrades by [X]%

## Data Requirements (ML-Specific)

### Training Data
THE SYSTEM SHALL require minimum [N] labeled examples
THE SYSTEM SHALL validate data quality before training

### Inference Data
WHEN input data is received
THE SYSTEM SHALL validate against expected schema
```

## Generating Kiro Spec

After creating PRD, generate the Kiro spec:

1. Create directory: `.kiro/specs/[product-slug]/`
2. Generate `requirements.md` with EARS-formatted requirements
3. Kiro will auto-generate `design.md` when you advance the spec
4. Kiro will auto-generate `tasks.md` from design

## Quality Checklist

Before completing:
- [ ] All PRFAQ features have corresponding requirements
- [ ] Requirements use EARS syntax correctly
- [ ] Acceptance criteria are testable
- [ ] Personas map to user stories
- [ ] Priorities are assigned (P0/P1/P2)
- [ ] ML requirements included (if applicable)
- [ ] Both PRD document and Kiro spec created