Write throwaway code to answer one falsifiable design question within a time-box.
npx skills add https://github.com/kieranpotts/skills --skill spike-kieranpotts
Please help me install this Agent Skill. Skill: spike Source: https://github.com/kieranpotts/skills/tree/main/skills/spike Command: npx skills add https://github.com/kieranpotts/skills --skill spike-kieranpotts
High-intent search queries and answers about installing and using this skill.
State one falsifiable question in a single sentence, define the evidence that would close it, and set a time-box before writing code. Then take the shortest path to that evidence with throwaway code, record the measured result, and file the verdict in the store that owns the question.
A spike answers one question and its code is always discarded, while a prototype is often intended to evolve into a product. Spike code deliberately omits tests, error handling, and configuration, so promoting it to production skips the specification and design discipline it was meant to inform.
A single spike's time-box should not exceed three to five days, and half a day or one day is typical. When the budget runs out, stop and report what was learned; extending the box is an explicit decision for the caller, and a question needing more than a week is really several spikes or project work.
No, spike code must never be promoted to production, however close to working it looks. What makes it cheap to write is everything it omits, so the capability is re-implemented cleanly from scratch using what the spike taught.
An inconclusive spike still reports what the budget bought and what remains unknown. It may propose another time-boxed spike that would disambiguate, or hand the decision back to the caller, but it never starts a second spike on its own initiative.
vertexcover-ioDocuments resolved implementation questions and decisions in qa.md before planning begins.
SgtPookiWrite ADRs, RFCs, design specs, and requirement documents with testable statements.
JSai23Prompt targeted architecture and design questions grounded in plans and documentation.
i-bebsiBuilds throwaway prototype experiments to validate technical feasibility before committing to a real build.
