What problem does it solve? It answers feasibility questions that research alone cannot resolve by building small, disposable prototypes, so you avoid committing to a full implementation before knowing whether an approach actually works. ## Core Features & Use Cases - Structured Spike Decomposition: Breaks an idea into 2-5 independent feasibility questions framed as Given/When/Then statements, ordered by risk so the idea-killing question runs first. - Comparison Spikes: Builds competing approaches side by side (e.g., pdfjs vs camelot for PDF parsing) and produces a head-to-head verdict table with a clear winner recommendation. - Evidence-Based Verdicts: Each spike directory closes with a VALIDATED, PARTIAL, or INVALIDATED verdict documenting what worked, what failed, and recommendations for the real build. - Use Case: Before building a real-time chat feature, spike whether WebSocket streaming delivers LLM tokens to the client under 100ms, then throw the prototype away once the question is answered. ## Quick Start Ask the agent to spike whether your chosen approach is feasible, for example: "Spike whether I can stream LLM responses over WebSockets with under 100ms latency before I commit to this architecture."
