What problem does it solve? Exploratory coding often drifts into unplanned implementation: no tests, no error handling, and code that quietly ships to production. This Skill enforces a disciplined prototype cycle so a single technical uncertainty gets answered cheaply without the throwaway code leaking into the real codebase. ## Core Features & Use Cases - Five-Step Protocol: Name one falsifiable question, set a timebox, build the minimum to answer it, record the answer where it survives the code, then delete the prototype. - Spike Branch Mapping: Aligns the prototype discipline with a spike/ branch convention that waives coverage and documentation gates while keeping linting, type checks, and secrets detection, with a two-week maximum lifespan and no merges to main. - Anti-Rationalization Guardrails: Red flags and pre-flight checks catch scope creep, bundled questions, and attempts to ship or harden prototype code instead of rewriting it on a feat/ branch. - Use Case: Before committing to a streaming architecture, run a timeboxed spike to verify a parser holds steady memory on a 2 GB file, record the throughput numbers, delete the branch, and start the real implementation with tests. ## Quick Start Ask the AI to run a timeboxed throwaway prototype on a spike branch that answers whether your chosen approach is viable, then record the answer and discard the code.
