recipe-define
Orchestrate PRD creation from validated hypotheses — standard PRD output with 4 Risks confidence and hypothesis traceability
Context: Transform validated hypotheses into a PRD with 4 Risks confidence scores, hypothesis traceability, and user stories. The PRD follows a standard structure that can be consumed by downstream implementation workflows (e.g., claude-code-workflows).
Orchestrator Definition
Execution Protocol:
- Delegate review to prd-reviewer (via Agent tool, subagent_type: "discover:prd-reviewer") for bias-free quality assessment
- Follow the definition flow defined below
- Stop at every
[STOP — BLOCKING]marker — present findings and CANNOT proceed until user explicitly confirms
Workflow Overview
Input (validated hypotheses + Opportunity context)
↓
1. Readiness Assessment → Check hypothesis confidence levels
↓
2. PRD Drafting → Using prd-template.md with discovery extensions
↓
3. User Story Generation → 4 Risks per story [Stop: User reviews PRD draft]
↓
4. Quality Review → prd-reviewer assesses the PRD [Stop: User reviews feedback]
↓
Output: docs/prd/[feature-name]-prd.md
Execution Decision Flow
1. Readiness Assessment
Input: $ARGUMENTS
Assess whether hypotheses are "validated enough" for PRD creation:
- Read relevant Opportunity and hypothesis files
- For each hypothesis intended for the PRD:
- Check confidence scores against thresholds (see prd-standards skill
references/user-story-guide.md) - Assess cost x risk x reversibility
- Determine: validated enough / needs more validation
- Check confidence scores against thresholds (see prd-standards skill
- See product-principles skill
references/mvp-definition.mdfor scope determination
Decision:
- All key hypotheses validated enough → Proceed to PRD drafting
- Some hypotheses below threshold → Present to user with options:
- Lower threshold (add risk mitigation like feature flags)
- Validate further (return to hypothesis validation)
- Proceed with documented remaining risks
2. PRD Drafting
Use prd-standards skill references/prd-template.md to create the PRD:
- Overview: Link to Opportunity and validated hypotheses
- User Stories: Apply 4 Risks assessment per story (see step 3)
- Functional Requirements: Derive from validated hypotheses with EARS-format ACs (see prd-standards skill
references/acceptance-criteria.md). Assign stable AC IDs (AC-001, AC-002, ...) sequentially across all requirements — IDs are global to the PRD, not reset per requirement - Design Context: Populate from project design artifacts:
- Design Principles: Copy from
docs/product/design-principles.md - Tone & Voice: Copy from
docs/product/design/brand-direction.md(if exists) - Design Guardrails: Derive Do's/Don'ts from design principles — specific patterns to follow, and anti-patterns with their preferred alternatives
- Visual Reference: Link to
brand-direction.mdand relevant prototype files
- Design Principles: Copy from
- Success Criteria: Tie to Product Outcomes from
docs/product/vision.md - Technical Dependencies: For each external API, library, or service, verify current availability and status with WebSearch before including in the PRD
- Assumptions (Unvalidated): Explicitly list hypotheses NOT yet validated that the PRD proceeds with
3. User Story Generation
For each user story, follow prd-standards skill references/user-story-guide.md:
- Write in persona-grounded format (As a [persona name], I want to...)
- Assess 4 Risks confidence with evidence
- Determine delivery readiness per story
- Document remaining risks per story
[STOP — BLOCKING] Present PRD draft to user for review:
- Complete PRD draft
- User stories with 4 Risks confidence summary
- Highlighted remaining risks and unvalidated assumptions
- MVP scope recommendation (Must Have / Should Have / Could Have)
CANNOT proceed to quality review until user explicitly approves the draft.
4. Quality Review
Invoke prd-reviewer using Agent tool (subagent_type: "discover:prd-reviewer") for bias-free assessment:
- prd-reviewer operates in a separate context
- It checks: completeness, internal consistency, evidence backing
- It identifies gaps the author might have missed
[STOP — BLOCKING] Present prd-reviewer feedback to user:
- prd-reviewer's assessment
- Required changes (if any)
- Approval recommendation
CANNOT finalize PRD until user explicitly approves after reviewing feedback.
If changes required:
- Apply changes to the PRD
- Re-run prd-reviewer if changes are significant
- Present updated PRD for final approval
5. File Output
After user approval:
- Write PRD to
docs/prd/[feature-name]-prd.md
Sub-agent Usage
| Agent | When | Why (context separation benefit) |
|---|---|---|
| prd-reviewer (subagent_type: "discover:prd-reviewer") | PRD quality review | Eliminates author's self-review bias |
PRD Structure
The PRD output follows a standard structure:
- Core sections: Overview, User Stories, Functional Requirements, Design Context, Non-Functional Requirements, Success Criteria, Technical Considerations
- Discovery extensions: Hypothesis & validation references, 4 Risks confidence per user story, design context (principles, tone, guardrails, visual reference), unvalidated assumptions section
- Storage:
docs/prd/
The extensions are additive — they don't replace or modify the core sections. This means the PRD can be consumed by any downstream workflow that expects the standard structure.
Scope Boundaries
Included: PRD creation, user story generation, quality review Not included: Hypothesis validation, Design Doc/ADR creation, implementation
Completion Criteria
- Hypothesis readiness assessed
- PRD drafted with hypothesis refs, 4 Risks, remaining risks
- User stories include 4 Risks confidence per story
- User reviewed PRD draft
- prd-reviewer assessed quality
- User approved final PRD
- PRD written to
docs/prd/