agentic-workflow-product-vision

name	agentic:workflow:product-vision
description	Use when creating a comprehensive product vision document from a rough idea. Triggers - new product initiative, strategic planning, need rich vision before specs, product direction definition.

/agentic:workflow:product-vision - Idea to Product Vision

Usage: /agentic:workflow:product-vision [<input>]

Transform a rough idea into a comprehensive product vision document through creative brainstorming and rigorous questioning. Always interactive — no auto mode.

Arguments

• No args: Prompt for idea
• path/to/notes.md: Use existing notes/idea file

Workflow Overview

1. Input Detection -> 2. Context Gathering (subagent) -> 3. Creative Exploration (subagent, brainstorming techniques) -> 4. Vision Discovery (subagent) -> 5. Confirm Vision [loop if rejected] -> 6. Vision Deepening -> 7. Write Vision Document

digraph workflow {
    rankdir=TB;
    input [label="Input Detection" shape=box];
    context [label="Context Gathering\n(subagent)" shape=box];
    brainstorm [label="Creative Exploration\n(subagent, brainstorming)" shape=box style=filled fillcolor="#e6f3ff"];
    discovery [label="Vision Discovery\n(subagent)" shape=box];
    confirm [label="Developer Confirms\nVision?" shape=diamond];
    refine [label="Re-run Discovery\nwith Feedback\n(subagent)" shape=box];
    deepen [label="Vision Deepening\n(one at a time)" shape=box];
    gate [label="Critical Open\nQuestions Remain?" shape=diamond];
    write [label="Write Vision Document\n(subagent + compile)" shape=box];
    done [label="DONE" shape=doublecircle];

    input -> context;
    context -> brainstorm;
    brainstorm -> discovery;
    discovery -> confirm;
    confirm -> deepen [label="confirmed"];
    confirm -> refine [label="rejected / corrections"];
    refine -> confirm;
    deepen -> gate;
    gate -> deepen [label="yes - keep asking"];
    gate -> write [label="no - ready"];
    write -> done;
}

---

Step Files

Execute steps in order. Each step file contains detailed instructions.

Step	File	Description
1	steps/step-01-input-detection.md	Parse args, init state, generate topic slug
2	steps/step-02-context-gathering.md	Gather product & technical context from codebase
3	steps/step-03-creative-exploration.md	Brainstorming session with creative techniques
4	steps/step-04-vision-discovery.md	Structured vision element discovery (informed by brainstorming)
5	steps/step-05-confirm-vision.md	Developer confirms vision direction; loop if rejected
6	steps/step-06-vision-deepening.md	Ask remaining vision questions one at a time
7	steps/step-07-write-vision.md	Compile comprehensive vision document

Start by reading steps/step-01-input-detection.md and follow NEXT STEP at end of each file.

---

Templates

Template	Description
templates/workflow-state.yaml	State tracking template
templates/vision-document.md	Output document template

Supporting Files

File	Description
brain-methods.csv	62 brainstorming techniques across 10 categories

Subagent References

All product work is delegated to the CPO subagent, which loads its own skills (product-vision, product-discovery, brainstorming) via mandatory setup.

Invoke: Task(subagent_type="general-purpose", prompt="You are the CPO agent. Read .claude/agents/agentic-agent-cpo.md for your full instructions and mandatory setup. ...")

---

Mandatory Delegation

You MUST delegate all discovery/brainstorming/context-gathering work using the Task tool. NEVER do it inline.

You are the orchestrator. You:

• Parse input
• Initialize state tracking
• Invoke subagents in sequence
• Handle handoffs between agents
• Validate outputs at each step
• Run the vision deepening loop yourself (step 6)
• Delegate vision document writing to CPO subagent

You NEVER:

• Gather codebase context yourself (delegate to Explore subagent)
• Run brainstorming sessions yourself (delegate to CPO subagent)
• Ask discovery questions yourself (delegate to CPO subagent)
• Write vision documents yourself (delegate to CPO subagent)

If you catch yourself doing agent work instead of delegating, STOP and use the Task tool.

---

THE GATE RULE

You MUST NOT proceed to step 7 (write vision) if critical open questions remain.

Critical = questions about vision direction, problem space, target users, strategic goals, value proposition, or product principles.

Minor = visual preferences, specific metric targets, implementation details. These can remain.

digraph gate {
    check [label="Any critical\nopen question?" shape=diamond];
    block [label="BLOCKED\nKeep asking in step 6" shape=box style=filled fillcolor="#ffcccc"];
    proceed [label="Proceed to step 7\nWrite vision" shape=box style=filled fillcolor="#ccffcc"];

    check -> block [label="yes"];
    check -> proceed [label="no"];
}

If the developer says "just decide" or "I don't care":

• Challenge once: "This shapes the vision. Are you sure you want me to choose?"
• If they insist: log as DEVELOPER_DEFERRED, make best-guess, proceed

---

Error Handling

Discovery Incomplete

If discovery output missing required sections:

1. Log which sections are missing
2. Ask user to provide missing info

Discovery Rejected Multiple Times

If developer rejects vision 3+ times:

1. Log specific complaints
2. Ask developer to describe what's wrong in their own words
3. Feed verbatim description to Discovery subagent

Developer Refuses to Answer Questions

If developer says "skip" or "later" on a critical question:

1. Explain why this question is critical for the vision
2. If they insist: mark as UNRESOLVED_CRITICAL
3. UNRESOLVED_CRITICAL questions block vision document generation

Step Failure

If any step fails:

1. Log error in workflow-state.yaml
2. Set status: "failed"
3. Present error, ask how to proceed

---

Artifacts

All outputs: .claude/_agentic_output/product/vision/{topic}/{instance_id}/

• workflow-state.yaml
• context-{topic}.md
• brainstorming-{topic}.md
• discovery-{topic}.md
• vision-decisions.md
• vision-{topic}.md