workflow-validate

name	workflow-validate
description	Validate technical feasibility with prototypes or tests. Use when uncertain about technical approach before committing to implementation.
disable-model-invocation	true

Technical Validation Command

User Input (Optional)

$ARGUMENTS

Optional input:

• Empty: Auto-determine validation method based on analysis.md
• Provided: Must consider user instructions (e.g., specific validation method, additional requirements)

---

Outline

1. Check Prerequisites:

   • Verify docs/work/WORK-{name}/analysis.md exists
   • If not, ERROR: "Run /workflow-analyze first"

2. Load Analysis Document:

   • Extract selected approach from analysis.md
   • Identify key technical assumptions and risks

3. Determine Validation Method:

   • Choose appropriate validation approach based on context:
     • Prototype implementation
     • UI/UX verification (Playwright MCP)
     • TDD approach
     • Library/API exploration
     • Technical documentation research
     • User verification delegation

4. Execute Validation:

   • Perform selected validation method
   • Create prototype code in __prototype__/{feature-name}/ if needed
   • Collect results and evidence

5. Assess Results:

   • Determine status: ✅ Success / ⚠️ Partial Success / ❌ Failure
   • Document findings and constraints

6. Write Documents (Dual Language):

   • Create docs/work/WORK-{name}/validation.ko.md (Korean - for user reference)
   • Create docs/work/WORK-{name}/validation.md (English - for agent consumption)
   • Include clear next steps recommendation

---

Key Rules

Documentation Language

CRITICAL: You must generate TWO versions of all documents:

1. Korean version (validation.ko.md): For user reference - written in Korean
2. English version (validation.md): For agent consumption - written in English

Both versions must contain identical structure and information, only the language differs.

Validation Principles

1. Focus on Core Risks: Validate only critical technical uncertainties
2. Practical Approach: Sufficient confidence over perfect validation
3. Clear Judgement: Explicit success/failure with next steps

Must Do

• Validate only what's uncertain
• Create minimal working code (if prototype)
• Document all findings clearly
• Provide actionable next steps
• Reference analysis.md for context
• Store prototype in __prototype__/ directory

Must Not Do

• Full implementation (save for execute phase)
• Validate obvious/known facts
• Repeat analysis.md content
• Give ambiguous conclusions ("maybe works")
• Over-engineer the validation

Validation Methods Selection

Choose based on uncertainty type:

Uncertainty Type	Validation Method	Output
Core logic feasibility	Prototype implementation	Working code + results
UI/UX changes	Playwright verification	Screenshots + test results
Complex algorithms	TDD approach	Test code + edge cases
External dependencies	Library exploration	Sample code + compatibility
Standards/patterns	Documentation research	Summary + recommendations
Environment-specific	User delegation	Test guide + checklist

Prototype Code Management

Location: __prototype__/{feature-name}/

How to create:

• Validate directly in actual codebase (modifying files is OK)
• After validation, extract core logic to __prototype__/
• Don't commit actual code changes (revert or leave as-is)

Purpose:

• Reference for plan/execute phases
• Archive of validated core logic
• Proof of technical feasibility

Lifecycle:

• Created during validate
• Referenced in execute
• User manages cleanup

---

Validation Status Guidelines

Success Criteria

• All core technical assumptions validated
• No blocking issues found
• Clear path to implementation
• Next: Proceed to /workflow-plan

Partial Success Criteria

• Main approach works with constraints
• Workarounds or alternatives available
• Trade-offs acceptable
• Next: Document constraints, get user confirmation

Failure Criteria

• Core approach not feasible
• Blocking issues without workarounds
• Fundamental assumptions invalid
• Next: Return to /workflow-analyze for re-evaluation

---

Spike Principles

This validation follows Agile Spike methodology:

1. Risk Reduction: Focus on highest risk items
2. Just Enough: Minimal code for maximum learning
3. Throwaway Code: Prototypes are for learning, not production
4. Clear Outcome: Binary decision on feasibility

---

Execution

Now start the validation task according to the guidelines above.