🔔 시스템 메시지: 이 Skill이 호출되면 [SEMO] Skill: explore-approach 호출 - {주제} 시스템 메시지를 첫 줄에 출력하세요.

explore-approach Skill

Purpose: Systematic technical exploration and evidence-based recommendation

When to Use

Agents should invoke this skill when:

• Multiple valid technical approaches exist
• Performance characteristics unknown
• Integration complexity unclear
• New technology/library evaluation needed
• plan.md indicates technical uncertainty

Quick Start

1. Create Spike Branch

git checkout -b spike/[topic-name]

2. Define Approaches (2-3)

Evaluation criteria:

• Performance (latency, throughput, resources)
• Complexity (LOC, learning curve)
• Maintainability (docs, community)
• Integration (stack compatibility)

3. Implement & Measure

• Time-box each prototype (1-2 hours max)
• Create spike-prototypes/[approach-name] directories
• Measure: setup time, LOC, dependencies, performance

4. Document Findings

Create docs/spikes/[topic-name].md with recommendation

Usage

skill: spike("realtime-tech-evaluation");
skill: spike("state-management", { focus: "performance" });
skill: spike("image-optimization", { quick: true });

Critical Rules

1. Spike Branches are Temporary: NEVER merge to main
2. Prototype, Don't Perfect: Minimal viable implementation
3. Measure Everything: Metrics mandatory
4. Evidence-Based Recommendations: NEVER recommend without data
5. Document for Future: Spike docs are permanent

Common Scenarios

| Scenario | Approaches | Duration | |----------|------------|----------| | Real-time | WebSocket vs SSE vs Polling | 2-3 hours | | State Management | Zustand vs Jotai vs Redux | 1-2 hours | | Image Optimization | next/image vs Cloudinary | 2-3 hours | | Authentication | NextAuth vs Supabase Auth | 3-4 hours |

Related Skills

• spec - Use recommendation in specification
• implement - Implement recommended approach
• verify - Validate implementation

References

For detailed documentation, see:

• Workflow - 8-phase spike process, cleanup
• Documentation Template - Spike doc structure, output format