AgentSkillsCN

Build One to Throw Away

在探索新概念、不熟悉的技术或不明确的需求时，应用一次性原型法。快速构建、深入学习，然后正确重建。用于可行性研究、概念验证和学习练习。

SKILL.md
--- frontmatter
name: "Build One to Throw Away"
description: "Apply throwaway prototyping methodology when exploring new concepts, unfamiliar technology, or unclear requirements. Build fast, learn deeply, then rebuild properly. Use for feasibility studies, proof-of-concepts, and learning exercises."
version: "2.0.0"

Build One to Throw Away

You are an expert in throwaway prototyping methodology, helping developers apply Fred Brooks' principle: "Plan to throw one away; you will, anyhow."

Core Principle

You learn by building. The first version reveals what you don't know. Instead of trying to perfect it, acknowledge it as a learning tool, capture the insights, and rebuild with confidence.

Philosophy

The best way to understand a problem is to solve it wrong first—intentionally, quickly, and with full awareness that the solution will be discarded. This isn't about sloppy work; it's about honest learning.

Key insight: The first attempt is research, not production. Treat it accordingly.

When to Use This Skill

USE when:

  • •Exploring unfamiliar technology or libraries
  • •Requirements are unclear or changing rapidly
  • •Feasibility is uncertain
  • •Architecture decisions need validation
  • •Team needs to learn domain or technology
  • •"Spike" or proof-of-concept is needed

DO NOT USE when:

  • •Requirements are clear and stable
  • •Technology is well-understood by the team
  • •Building something simple or familiar
  • •Time constraints don't allow for rebuild
  • •After the fourth iteration (Brooks' warning: stop rebuilding)

The Three-Phase Process

Phase 1: Fast Prototype (Learning Mode)

Goal: Learn as much as possible, as quickly as possible

Skip: Tests, error handling, edge cases, documentation, performance optimization, design patterns Focus: Core logic, happy path, discovery, experimentation

Detailed guidance: phase-details.md

Phase 2: Extract the Learning (Documentation)

Goal: Capture insights before they fade

Document: What you learned, what surprised you, what didn't work, architecture decisions, key discoveries Create: Learning log, ADRs, transition document with recommendations

Detailed guidance: phase-details.md

Phase 3: Proper Build (Production Mode)

Goal: Build it right using what you learned

Apply: TDD, proper architecture, error handling, documentation, design patterns Build with: Confidence, clarity, and understanding gained from Phase 1

Detailed guidance: phase-details.md

Key Anti-Patterns (Avoid These)

  1. •"We'll just clean up the prototype" - Don't. Rebuild from scratch.
  2. •"Second-system effect" - Don't add every possible feature you discovered.
  3. •"Throw away two" - Same team must build both versions.
  4. •"Perfect prototype" - Speed over quality in Phase 1.
  5. •"Documentation-free" - Must capture learnings in Phase 2.

Full explanations and solutions: anti-patterns.md

Available Resources

Core Documentation

  • •

    phase-details.md - Comprehensive guide for each phase

    • •What to skip/focus on in prototyping
    • •Learning log templates
    • •Transition procedures
    • •Read when: Executing a specific phase
  • •

    implementation-guide.md - Step-by-step instructions

    • •Decision checklist
    • •Project setup commands
    • •README and learning log templates
    • •Transition document structure
    • •Read when: Starting a new throwaway prototype
  • •

    examples.md - Real-world case studies

    • •API integration example
    • •Complex algorithm example
    • •New technology exploration
    • •Before/after code comparisons
    • •Read when: Want to see the pattern in action

Advanced Topics

  • •

    historical-context.md - Fred Brooks and the principle's origins

    • •The Mythical Man-Month background
    • •OS/360 project experience
    • •Evolution and refinement
    • •Cultural challenges
    • •Read when: Want deeper understanding of the methodology
  • •

    anti-patterns.md - Detailed mistake prevention

    • •Five major anti-patterns explained
    • •Warning signs and recognition
    • •Solutions and recovery strategies
    • •Read when: Planning or reviewing throwaway work

Quick Start

bash
# Create throwaway directory
mkdir throwaway-prototype
cd throwaway-prototype

# Mark it clearly as throwaway
cat > README.md << 'EOF'
# ⚠️ THROWAWAY PROTOTYPE - DO NOT USE IN PRODUCTION

**Purpose:** [describe what you're learning]
**To be discarded:** [target date]
**Learning goal:** [specific questions to answer]

This is intentionally incomplete and will be rebuilt properly.
EOF

# Start coding with freedom to explore

Complete setup guide: implementation-guide.md

Success Criteria

✅ Prototype answered your key questions ✅ Learning documented before rebuilding ✅ Architecture decisions recorded (ADRs) ✅ Proper version built with TDD ✅ Prototype code deleted (not evolved)

Typical Workflow

  1. •Identify the unknown - What needs discovery?
  2. •Fast prototype - Build to learn, not to last
  3. •Document insights - Capture before rebuilding (use phase-details.md)
  4. •Delete prototype - Commit to rebuild
  5. •Proper build - Apply learnings with quality practices

Key Principles

  • •Prototype with purpose - Know what you're trying to learn
  • •Speed over quality - In Phase 1 only
  • •Document insights - Phase 2 is mandatory, not optional
  • •Rebuild, don't refactor - Start fresh in Phase 3
  • •One prototype, one rebuild - Don't iterate indefinitely

Remember

The throwaway prototype is temporary scaffolding—it helps you build, but it's not part of the structure. The value isn't in the code; it's in the knowledge gained. Brooks' wisdom: "The management question, therefore, is not whether to build a pilot system and throw it away. You will do that. The only question is whether to plan in advance to build a throwaway, or to promise to deliver the throwaway to customers."

Build intentionally imperfect code, learn intentionally, then build it right.

For comprehensive guidance, explore the reference/ directory based on your current need.