Imported from sakuradairong/omp-config (skills/spike/SKILL.md). Install upstream with npx skills add sakuradairong/omp-config --skill spike. Copyright stays with the author.

Spike — Throwaway Experiments

When to use:

• Validating an approach before committing to it
• Exploring unfamiliar APIs, libraries, or patterns
• Testing performance characteristics
• Confirming whether a design works

Procedure

1. Define the Question

Write down exactly what you need to learn:

• "Does library X support feature Y?"
• "Will approach Z achieve acceptable performance?"
• "Is pattern W feasible in this codebase?"

2. Build Minimal Experiment

• Create isolated prototype (new file/scratch directory)
• Minimal code to answer the question — nothing else
• Hardcode inputs, skip error handling, no tests
• This is throwaway code — quality doesn't matter

3. Run and Learn

• Execute the experiment
• Collect data, timings, outputs
• Document the answer to your question

4. Decide

• Proceed: Approach validated → delete spike code, implement properly
• Pivot: Approach doesn't work → try alternative
• Abandon: Feature not feasible → document and move on

Rules

• Spike code is throwaway. Never evolve it into production code.
• Timebox. Set a timer (30-60 min). If not answered, simplify question.
• Delete when done. Keep only the learnings, not the code.
• Document decision. Write down what you learned and why.