PRD Creator

Philosophy

We document for alignment, not approval. A PRD answers:

1. What are we building and why?
2. Who is it for?
3. How do we know it worked?
4. What's the smallest thing we can ship?
5. Any compliance blockers?

Workflow Modes

Discovery Mode

When PM has a rough idea, guide them through conversationally:

1. Problem — "What's the user pain or opportunity?"
2. Evidence — "How do you know this matters?" (even anecdotal is fine)
3. Why now? — "Why this over other things we could build? What's the cost of waiting?"
4. Persona — "Who specifically has this problem?"
5. Success — "If this works, what changes? How will we measure it?"
6. First slice — "What's the smallest version that tests the hypothesis?"
7. Risks — "Any compliance, security, or regulatory concerns? Any technical blockers?"

Don't ask all at once. Have a conversation. Fill gaps with reasonable assumptions and flag them as open questions.

See references/examples.md for a sample Discovery conversation.

Draft Mode

When ready to write, use assets/prd-template.md. The template includes:

• Overview — What the feature does and why it matters
• Background — Context, research, and strategic rationale
• Goals & Success Metrics — Measurable outcomes and targets
• User Stories — Stories with US-### IDs in As a / I want / So that format
• Requirements — Functional and non-functional requirements
• User Experience — Key workflows, edge cases, and error states
• Technical Considerations — Dependencies, constraints, architecture
• Acceptance Criteria — Testable conditions for completion
• Open Questions — Unresolved items with Q-### IDs
• Out of Scope — What's explicitly excluded
• Timeline & Milestones — Key dates and phases
• Risks & Mitigations — What could go wrong and how to address it

Story Expansion Mode

Use this mode to add acceptance criteria to stories before exporting to Jira. The PM workflow is:

1. Draft PRD → Write stories in user story format (As a / I want / so that)
2. Expand stories → Add acceptance criteria using this mode
3. Export to Jira → Use jira-prd-export to create epics/stories with full ACs

To expand a story:

1. Take an existing user story (e.g., US-001)
2. Generate acceptance criteria in Given/When/Then format
3. Use AC-###.# IDs (e.g., AC-001.1, AC-001.2)
4. Reference references/story-patterns.md for common patterns

See references/examples.md for story expansion examples.

Epic Generation Mode

When organizing stories into epics:

• 1 epic = 1-2 sprints of work max
• Group related user stories under a theme
• Each epic gets a clear objective and success criteria

See references/examples.md for epic format example.

Interacting with Codebase

When PM asks about existing features or feasibility:

1. Search codebase for relevant files
2. Summarize what exists in plain language
3. Note integration points or constraints
4. Don't over-engineer — flag technical questions for eng

Output Formats

Need	Output
Quick alignment	Markdown in chat, no file needed
Sprint planning	PRD file → confluence-prd-export → jira-prd-export
Symphony pipeline	PRD file → prd-analyst → requirements.json

ID Conventions

Consistent IDs enable traceability through the Symphony pipeline:

Type	Format	Example
User Story	US-###	US-001, US-002
Acceptance Criteria	AC-###.#	AC-001.1, AC-001.2
Open Question	Q-###	Q-001, Q-002

Integration with Symphony

PRDs created with this skill feed into Symphony's planning pipeline:

prd-creator → PRD.md → prd-analyst → requirements.json → plan-writer → implementation-plan.md

The prd-analyst agent extracts:

• User stories → user_stories[]
• Acceptance criteria → acceptance_criteria[]
• Success metrics → success_metrics[]
• Open questions → open_questions[]
• Risks and constraints → constraints[]

Task Tracking

Use TodoWrite to track progress through each workflow mode.

Discovery Mode

TodoWrite([
  {"content": "Understand the problem and user pain", "status": "pending", "activeForm": "Understanding the problem"},
  {"content": "Gather evidence for why this matters", "status": "pending", "activeForm": "Gathering evidence"},
  {"content": "Identify target persona(s)", "status": "pending", "activeForm": "Identifying personas"},
  {"content": "Define success metrics", "status": "pending", "activeForm": "Defining success metrics"},
  {"content": "Scope the first slice / MVP", "status": "pending", "activeForm": "Scoping MVP"},
  {"content": "Identify risks and compliance concerns", "status": "pending", "activeForm": "Identifying risks"}
])

Draft Mode

TodoWrite([
  {"content": "Write Overview section", "status": "pending", "activeForm": "Writing Overview"},
  {"content": "Write Background section", "status": "pending", "activeForm": "Writing Background"},
  {"content": "Write Goals & Success Metrics", "status": "pending", "activeForm": "Writing Goals & Success Metrics"},
  {"content": "Write User Stories with US-### IDs", "status": "pending", "activeForm": "Writing User Stories"},
  {"content": "Write Requirements (Functional + Non-Functional)", "status": "pending", "activeForm": "Writing Requirements"},
  {"content": "Write User Experience (Workflows, Edge Cases)", "status": "pending", "activeForm": "Writing User Experience"},
  {"content": "Write Technical Considerations", "status": "pending", "activeForm": "Writing Technical Considerations"},
  {"content": "Write Acceptance Criteria", "status": "pending", "activeForm": "Writing Acceptance Criteria"},
  {"content": "Document Open Questions with Q-### IDs", "status": "pending", "activeForm": "Documenting questions"},
  {"content": "Define Out of Scope", "status": "pending", "activeForm": "Defining Out of Scope"},
  {"content": "Write Risks & Mitigations", "status": "pending", "activeForm": "Writing Risks & Mitigations"}
])

Story Expansion Mode

TodoWrite([
  {"content": "Select user story to expand", "status": "pending", "activeForm": "Selecting story"},
  {"content": "Generate acceptance criteria (Given/When/Then)", "status": "pending", "activeForm": "Generating acceptance criteria"},
  {"content": "Assign AC-###.# IDs", "status": "pending", "activeForm": "Assigning AC IDs"},
  {"content": "Review against story-patterns.md", "status": "pending", "activeForm": "Reviewing patterns"}
])

Mark each task completed only after the PM confirms alignment. For Discovery Mode, tasks may be revisited as the conversation evolves.

Keep It Light

• Keep to as few pages as is feasible
• Assumptions are fine — flag them as Q-### items
• Details emerge in Story Expansion Mode
• The AI can regenerate/expand sections later
• Perfect is the enemy of shipped