name: write-prd license: MIT compatibility: "Claude Code 2.1.170+." description: "Write PRD — Product Requirements Documents with structured 8-section templates, user stories, acceptance criteria, and value proposition validation. Use when writing PRDs, defining product requirements, creating user stories with INVEST criteria, or building go/no-go decision frameworks." tags: [prd, requirements, user-story, acceptance-criteria, invest, value-proposition, go-no-go] context: fork agent: product-strategist version: 2.0.1 author: OrchestKit user-invocable: true argument-hint: "product name or feature" disable-model-invocation: false complexity: medium persuasion-type: collaborative metadata: category: document-asset-creation allowed-tools:
- Read
- Write
- Edit
- Glob
- Grep
- Bash
- WebFetch
- WebSearch
- AskUserQuestion
- TaskCreate
- TaskUpdate
- Agent
- mcp__memory__search_nodes
- mcp__memory__create_entities
- mcp__memory__add_observations
- mcp__memory__create_relations
triggers:
keywords: [prd, "product requirements", "requirements document", "requirements for", "define the requirements", "user stories", "acceptance criteria", "product spec", "formal spec"]
examples:
- "write a PRD for the SSO invite flow"
- "create a product requirements document for billing"
- "i need user stories and acceptance criteria for the dashboard" anti-triggers: [implement, build, brainstorm, explore, review]
PRD — Product Requirements Document
Translate product vision and research into clear, actionable engineering specifications. Produces PRD-[product-name].md output files following an 8-section structure.
Output file naming: PRD-[product-name].md (e.g., PRD-sso-invite-flow.md)
Argument Resolution
PRODUCT = "$ARGUMENTS" # Product name or feature, e.g., "SSO invite flow"
STEP 0: Scope Clarification
AskUserQuestion(
questions=[{
"question": "What type of PRD?",
"header": "PRD Scope",
"options": [
{"label": "Full PRD (Recommended)", "description": "All 8 sections with research, stories, and release plan"},
{"label": "Lightweight spec", "description": "Summary, objectives, user stories only"},
{"label": "User stories only", "description": "INVEST stories with acceptance criteria"},
{"label": "Update existing PRD", "description": "I have a PRD file to iterate on"}
],
"multiSelect": false
}]
)
Task Management
# 1. Create main task IMMEDIATELY
TaskCreate(subject="Write PRD: {PRODUCT}", description="8-section PRD with user stories and acceptance criteria", activeForm="Writing PRD for {PRODUCT}")
# 2. Create subtasks for each phase
TaskCreate(subject="Scope clarification", activeForm="Clarifying PRD scope") # id=2
TaskCreate(subject="Research and memory check", activeForm="Researching prior PRDs") # id=3
TaskCreate(subject="Draft 8-section PRD", activeForm="Drafting PRD sections") # id=4
TaskCreate(subject="Write user stories and acceptance criteria", activeForm="Writing user stories") # id=5
TaskCreate(subject="Write output file", activeForm="Writing PRD file") # id=6
# 3. Set dependencies for sequential phases
TaskUpdate(taskId="3", addBlockedBy=["2"]) # Research needs scope first
TaskUpdate(taskId="4", addBlockedBy=["3"]) # Drafting needs research context
TaskUpdate(taskId="5", addBlockedBy=["4"]) # Stories need draft structure
TaskUpdate(taskId="6", addBlockedBy=["5"]) # Output needs all sections done
# 4. Before starting each task, verify it's unblocked
task = TaskGet(taskId="2") # Verify blockedBy is empty
# 5. Update status as you progress
TaskUpdate(taskId="2", status="in_progress") # When starting
TaskUpdate(taskId="2", status="completed") # When done — repeat for each subtask
Memory Integration
# Search for prior PRDs and product decisions
mcp__memory__search_nodes(query="{PRODUCT} PRD requirements")
# After PRD is written, store key decisions
mcp__memory__create_entities(entities=[{
"name": "PRD-{product-slug}",
"entityType": "document",
"observations": ["PRD written for {PRODUCT}", "Key objectives: ..."]
}])
The 8-Section PRD Template
Load Read("${CLAUDE_SKILL_DIR}/references/prd-template.md") for the full template with all 8 sections (Summary, Contacts, Background, Objective, Market Segments, Value Propositions, Solution, Release), priority levels, and NFR categories.
User Stories & Acceptance Criteria
Load Read("${CLAUDE_SKILL_DIR}/references/user-stories-guide.md") for INVEST criteria, story format, Gherkin acceptance criteria, and Definition of Ready/Done.
Value Proposition Canvas
Load Read("${CLAUDE_SKILL_DIR}/references/value-prop-canvas-guide.md") for the canvas template and fit check process. Every Value Map item must correspond to a Job, Pain, or Gain.
Go/No-Go Gate Criteria
Load from rules: Read("${CLAUDE_SKILL_DIR}/rules/strategy-go-no-go.md") for stage gate criteria and scoring thresholds (Go >= 7.0 | Conditional 5.0-6.9 | No-Go < 5.0).
Rules (Load On-Demand)
- research-requirements-prd.md — INVEST user stories, PRD template, priority levels, DoR/DoD
- strategy-value-prop.md — Value proposition canvas, JTBD framework, fit assessment
- strategy-go-no-go.md — Stage gate criteria, scoring, build/buy/partner decision matrix
References
- output-templates.md — Structured JSON output schemas for PRD, business case, and strategy artifacts
- value-prop-canvas-guide.md — Detailed value proposition canvas facilitation guide
Output
After generating the PRD, write it to disk:
Write(f"PRD-{product_slug}.md", prd_content)
TaskUpdate(status="completed")
Plan-mode filenames (CC 2.1.111+): If this skill enters plan mode (via
EnterPlanMode) before writing the PRD, CC 2.1.111 names the resulting plan file after the prompt stem rather than random words. To get a clean filename, include the product slug in the plan-mode prompt (e.g.EnterPlanMode("PRD: authentication-rework")→plans/prd-authentication-rework.md). Before 2.1.111, plan filenames wereplans/swift-cobra.mdetc. — review and rename as needed.
Chain: Next Steps
After PRD is approved, decompose its acceptance criteria into a runnable goal, then implement:
/ork:prd-to-goal PRD-{product-slug}.md # → copy-pasteable `/goal until … abort-if …` line
/ork:implement PRD-{product-slug}.md
/ork:prd-to-goal reduces the Go/No-Go gate + acceptance criteria below into a single AND-joined boolean assertion, so the PRD drives an autonomous /goal run rather than a manual read-through.
Related Skills
ork:user-research— Build user understanding (personas, journey maps, interviews) before writing the PRDork:prd-to-goal— Reduce this PRD's acceptance criteria to a/goal until … abort-if …lineork:implement— Execute the implementation plan from the PRDork:brainstorm— Explore solution alternatives before committing to PRD scopeork:assess— Rate PRD quality and completeness
Version: 2.0.0