spike

What problem does it solve? When an implementation idea carries real uncertainty, reading code or docs often cannot resolve it. This Skill structures a small, disposable experiment (a spike) so you can validate or reject the idea quickly without touching production code paths. ## Core Features & Use Cases - Hypothesis-Driven Method: Forces you to state the hypothesis, the risk it validates, and a small observable result before writing any code. - Safety Guardrails: Keeps spikes local and disposable, prefers fixtures or mocked data, and forbids real tenant data, RLS/auth bypasses, or unapproved dependencies. - Verdict Recording: Ends every spike with a clear outcome — keep, reject, or investigate further — and requires throwaway code to be deleted or isolated. - Use Case: You are unsure whether a BullMQ retry strategy will handle a specific failure pattern in the EVA backend. Run a spike with mocked queues to observe the behavior, record the verdict, and only then convert the validated approach into tested production code. ## Quick Start Ask the agent to run a spike that validates whether your proposed approach works, using local fixtures and without modifying production code.