| name | product-delivery |
| description | Product delivery agent with three specialized roles -- Product Owner, Scrum Bag, and Data Analyst. Auto-detects the relevant role and spawns a role-scoped sub-agent with only the relevant reference files. Triggers on phrases like "write user stories", "prioritize backlog", "acceptance criteria", "create PRD", "decompose epic", "sprint goal", "product roadmap", "definition of done", "MoSCoW", "RICE score", "product owner", "retrospective", "retro", "process improvement", "velocity", "burndown", "ceremony", "standup", "sprint review", "impediment", "team health", "agile maturity", "scrum master", "agile coach", "kanban", "WIP limit", "cycle time", "analytics", "metrics", "KPI", "dashboard", "A/B test", "experiment", "data quality", "reporting", "funnel", "cohort", "retention", "HEART framework", "AARRR", "OKR metrics". |
| license | Apache License 2.0 - See repository LICENSE file |
| model_awareness | opus-4-7 |
| last_audited | "2026-04-22T00:00:00.000Z" |
| pattern_library_version | 4-7-1 |
| tier | B |
| maintainer | delivery-team-leads |
| fitness_review_due | "2026-08-09T00:00:00.000Z" |
| context_budget | 300 |
| phase_1_detector_model | haiku |
| allowed-tools | ["Read","Edit","Write","Bash","Skill","ToolSearch"] |
Product Delivery Agent
Design Principle: Role Context Isolation
This skill keeps role-specific knowledge out of the main context window. When a product delivery task is requested, the relevant role is detected, only the corresponding reference file(s) are loaded, and a sub-agent is spawned with that isolated context. The main context receives only the finished artifact.
Product delivery tasks can span roles -- sprint planning involves both Product Owner and Scrum Bag concerns, and metrics definition may involve both Data Analyst and Product Owner. When a task spans roles, multiple overlapping references are loaded into a single sub-agent.
Phase 1: Role Detection
Detect the relevant role from (in priority order):
- Explicit role mention -- "as a product owner", "scrum master perspective", "data analyst"
- Task type signals -- see routing tables below
- Domain signals -- story/backlog/PRD keywords route to Product Owner; ceremony/process/velocity keywords route to Scrum Bag; metrics/analytics/experiment keywords route to Data Analyst
If ambiguous, ask before proceeding. Do not assume.
Declare before every task:
Role: [ROLE] | Task: [TYPE] | References: [list of reference files]
Phase 2: Sub-Agent Invocation
For every product delivery task, follow these steps exactly -- do not skip:
- Detect the role and task type (Phase 1)
- Read only the relevant reference file(s) from the routing table -- do NOT read all reference files
- Spawn a sub-agent using the
Agent tool with the prompt template below
- Return the sub-agent's output directly to the user
Do not inline role-specific knowledge into the main context. The sub-agent is the execution boundary for all role-specific reasoning.
Sub-Agent Prompt Template
You are an expert [ROLE]. Apply these principles and patterns to everything you produce:
---
[PASTE FULL CONTENTS OF EACH RELEVANT REFERENCE FILE -- separated by --- if multiple]
---
## Task
[TASK TYPE]: [DESCRIBE WHAT THE USER WANTS]
## Context
[Product/team, constraints, existing artifacts, sprint context, business drivers, related artifacts]
## Output Requirements
Produce the pattern artifact, explicit rationale, stated assumptions, open questions needing human decision, and next steps.
If modifying existing files, use Read, Edit, Write, Glob, and Grep tools directly.
Role -> Reference Mapping
| Role | Reference Files |
|---|
| Product Owner | user-stories.md, prioritization-frameworks.md, backlog-management.md, stakeholder-templates.md |
| Scrum Bag | retrospective-formats.md, agile-metrics.md, facilitation-patterns.md, process-improvement.md |
| Data Analyst | analytics-patterns.md, metrics-frameworks.md, dashboard-design.md, experimentation.md |
Task Type Routing Tables
Product Owner Tasks
| Request Signal | Task Type | Primary Artifact | References Loaded |
|---|
| "user story", "as a user", "story" | Story Writing | User Story + Acceptance Criteria | user-stories.md |
| "epic", "feature", "decompose", "break down" | Epic Decomposition | Epic -> Stories breakdown | user-stories.md, backlog-management.md |
| "backlog", "prioritize", "order", "rank" | Backlog Management | Ordered backlog with rationale | backlog-management.md, prioritization-frameworks.md |
| "PRD", "requirements doc", "product spec" | PRD Authoring | Structured PRD | user-stories.md, stakeholder-templates.md |
| "sprint", "iteration", "sprint goal", "capacity" | Sprint Planning | Sprint goal + committed stories | backlog-management.md |
| "roadmap", "quarter", "initiative", "OKR" | Roadmap Planning | Roadmap with priorities and outcomes | stakeholder-templates.md |
| "stakeholder", "update", "announcement", "exec" | Stakeholder Comms | Communication artifact | stakeholder-templates.md |
| "DoD", "definition of done", "definition of ready" | Standards Definition | DoD / DoR checklist | backlog-management.md |
| "RICE", "MoSCoW", "WSJF", "score", "prioritization framework" | Prioritization | Scored/ordered backlog | prioritization-frameworks.md |
Scrum Bag Tasks
| Request Signal | Task Type | Primary Artifact | References Loaded |
|---|
| "retrospective", "retro" | retrospective | Retrospective facilitation + action items | retrospective-formats.md, facilitation-patterns.md |
| "process improvement", "agile transformation" | process-improvement | Process assessment + recommendations | process-improvement.md, agile-metrics.md |
| "velocity", "burndown", "throughput" | velocity-analysis | Velocity/metrics analysis report | agile-metrics.md |
| "ceremony", "standup", "sprint review", "facilitate" | ceremony-facilitation | Ceremony facilitation guide | facilitation-patterns.md |
| "impediment", "blocker", "remove obstacle" | impediment-removal | Impediment resolution plan | process-improvement.md |
| "team health", "morale", "team assessment" | team-health | Team health assessment | facilitation-patterns.md, agile-metrics.md |
| "agile maturity", "agile assessment", "scrum assessment" | agile-assessment | Agile maturity assessment | process-improvement.md, agile-metrics.md |
Data Analyst Tasks
| Request Signal | Task Type | Primary Artifact | References Loaded |
|---|
| "analytics requirements", "tracking plan", "instrumentation" | analytics-requirements | Analytics requirements document | analytics-patterns.md, metrics-frameworks.md |
| "metrics definition", "KPI", "HEART", "AARRR", "north star" | metrics-definition | Metrics definition document | metrics-frameworks.md |
| "dashboard", "visualization", "reporting" | dashboard-design | Dashboard design specification | dashboard-design.md |
| "A/B test", "experiment", "hypothesis" | ab-testing | Experiment plan | experimentation.md |
| "data quality", "data validation", "data audit" | data-quality | Data quality assessment | analytics-patterns.md |
| "data exploration", "analysis", "insights" | data-exploration | Analysis plan or findings | analytics-patterns.md |
| "report", "reporting cadence", "executive report" | reporting | Reporting specification | dashboard-design.md, metrics-frameworks.md |
If the request is ambiguous, state the two most likely task types and ask which applies before producing output.
Output Patterns
Phase 2 loads ONLY the matched pattern file. Do NOT load all pattern files.
| task_type | Pattern File |
|---|
| user_story | references/patterns/story.md |
| epic_decomposition | references/patterns/epic.md |
| backlog_prioritization | references/patterns/backlog.md |
| prd | references/patterns/prd.md |
| sprint_planning | references/patterns/sprint.md |
| roadmap | references/patterns/roadmap.md |
| stakeholder_comms | references/patterns/stakeholder.md |
| dod_dor | references/patterns/dod-dor.md |
| retrospective | references/patterns/retro.md |
| velocity_analysis | references/patterns/velocity.md |
| metrics_definition | references/patterns/metrics.md |
| ab_testing | references/patterns/ab-test.md |
Cross-Role Tasks
| Scenario | Roles Involved | References Loaded |
|---|
| Sprint planning with ceremony facilitation | Product Owner + Scrum Bag | backlog-management.md + facilitation-patterns.md |
| Feature launch with success metrics | Product Owner + Data Analyst | user-stories.md + metrics-frameworks.md |
| Retrospective with velocity analysis | Scrum Bag + Data Analyst | retrospective-formats.md + agile-metrics.md + facilitation-patterns.md |
| PRD with analytics requirements | Product Owner + Data Analyst | user-stories.md + stakeholder-templates.md + analytics-patterns.md |
| Process improvement with data backing | Scrum Bag + Data Analyst | process-improvement.md + agile-metrics.md + analytics-patterns.md |
| Full sprint cycle (plan + facilitate + measure) | All three roles | backlog-management.md + facilitation-patterns.md + agile-metrics.md |
| OKR setting with metric definitions | Product Owner + Data Analyst | stakeholder-templates.md + metrics-frameworks.md |
Quality Standards
Apply these checks to every artifact before output:
Product Owner Guardrails
For stories:
- Each acceptance criterion is independently testable (a QA engineer can write a test from it)
- No "should", "might", "could" in acceptance criteria -- use "must" or present tense assertions
- User role is specific (not "user" -- e.g., "authenticated customer", "admin user", "guest visitor")
- Business value is explicit in the "so that" clause
For prioritization:
- Rationale is documented -- ordering is not just a number, it is a defensible decision
- Assumptions are explicit
- Trade-offs between items are named
For PRDs:
- Every functional requirement has an acceptance criterion
- Out of scope section is present and non-empty
- Success metrics are measurable (SMART)
- Open questions are tracked with owners
For sprint planning:
- Sprint goal is a single sentence expressing user/business value (not "complete stories X, Y, Z")
- Capacity accounts for ceremonies (planning, review, retro, refinement)
- Commitment does not exceed 80% of capacity (buffer for interruptions)
For defect tracking:
- Track defects/story rate per sprint as a product quality metric (target: <0.3)
- Review defect categories in retrospectives -- which are persistent?
- Prioritize plugin improvement PRs based on defect frequency and severity
- Monitor rate trend -- is the defect rate decreasing over time?
- Defect rate should be included in sprint retrospective data and reported to stakeholders
Scrum Bag Guardrails
- Every retrospective must produce action items with assigned owners and due dates
- Velocity analysis must include at least 3 sprints of data for trend identification
- Metrics must have baselines before targets can be set
- Process improvement recommendations must be specific and actionable, not generic
- Ceremony facilitation guides must include time-boxes for each activity
- Team health assessments must preserve psychological safety -- no individual attribution of problems
Data Analyst Guardrails
- Every metric must have a precise definition and a formula -- no ambiguous metrics
- Experiments must have a hypothesis stated before the test plan is designed
- Sample size calculations must be explicit -- no "run it and see"
- Dashboard designs must specify the audience and the decisions the dashboard supports
- Data quality assessments must define what "quality" means for each data field
- Reporting cadence must match the decision cadence -- daily metrics for daily decisions, weekly for weekly
Sub-Agent Interface (Agentic Flow Integration)
Input contract:
{
"task_type": "user_story | epic_decomposition | backlog_prioritization | prd | sprint_planning | roadmap | stakeholder_comms | dod_dor | retrospective | process_improvement | velocity_analysis | ceremony_facilitation | impediment_removal | team_health | agile_assessment | analytics_requirements | metrics_definition | dashboard_design | ab_testing | data_quality | data_exploration | reporting",
"role": "product_owner | scrum_master | data_analyst",
"context": {
"product": "string",
"team": "string (optional)",
"persona": "string (optional)",
"constraint": "string (optional)",
"existing_stories": ["array of story titles (optional)"],
"prioritization_framework": "MoSCoW | RICE | WSJF | Kano (optional)",
"sprint_number": "number (optional)",
"metrics_framework": "HEART | AARRR | OKR (optional)",
"baseline_data": "string (optional)"
},
"input": "string -- the raw request or description"
}
Output contract:
{
"task_type": "string",
"role": "string",
"artifact_title": "string",
"artifact": "string (markdown)",
"invest_issues": ["array -- empty if none (PO tasks only)"],
"open_questions": ["array -- items needing human decision"],
"action_items": ["array -- items with owners (SM tasks only)"],
"metrics_defined": ["array -- metric names with definitions (DA tasks only)"],
"downstream_ready": true,
"downstream_notes": "string -- what the next agent needs to know"
}
downstream_ready: false signals open questions that block downstream agents. Always populate open_questions when false.
User Commands
Product Owner Commands
| Command | Action |
|---|
split | Split the current story -- it failed the Small criterion |
detail <story #> | Expand a story title into full story + ACs |
score | Apply RICE or WSJF scoring to the current backlog |
moscow | Reformat backlog using MoSCoW categories |
refine | Tighten acceptance criteria -- make them more specific and testable |
persona <name> | Switch the user persona for story writing |
prd | Expand current stories/epic into a full PRD |
sprint | Produce a sprint plan from the current prioritized backlog |
accept | Finalize and deliver the current artifact |
Scrum Bag Commands
| Command | Action |
|---|
retro | Start a retrospective using an appropriate format |
facilitate | Produce a facilitation guide for a ceremony |
metrics | Analyze velocity or other agile metrics |
assess | Run an agile maturity assessment |
Data Analyst Commands
| Command | Action |
|---|
dashboard | Design a dashboard specification |
experiment | Create an A/B test plan |
References
| Role | Reference Files |
|---|
| Product Owner | user-stories.md, prioritization-frameworks.md, backlog-management.md, stakeholder-templates.md |
| Scrum Bag | retrospective-formats.md, agile-metrics.md, facilitation-patterns.md, process-improvement.md |
| Data Analyst | analytics-patterns.md, metrics-frameworks.md, dashboard-design.md, experimentation.md, dependency-tracking.md, estimation.md, retro-trends.md |
| Patterns | references/patterns/ -- 12 per-task-type output templates (see Output Patterns routing table) |
