spike

What problem does it solve? Teams often commit to building features without knowing whether the underlying approach actually works. This Skill structures quick, disposable experiments (spikes) that answer feasibility questions with real running code instead of speculation, so bad ideas fail cheaply. ## Core Features & Use Cases - Decompose ideas into spike questions: Breaks a vague idea into 2-5 independent Given/When/Then feasibility questions, ordered by risk so the idea-killing question runs first. - Guided research and build loop: Researches competing libraries with web search and doc extraction, then builds a minimal runnable CLI, HTML page, or test per spike in its own spikes/NNN-name/ directory. - Comparison spikes and verdicts: Runs parallel head-to-head approaches (e.g., pdfjs vs camelot) and closes each spike with a VALIDATED, PARTIAL, or INVALIDATED verdict plus a recommendation for the real build. - Use Case: Before committing to a websocket-based streaming architecture, spike whether an LLM token stream reaches the client in under 100ms, and compare two PDF parsing libraries side by side to pick a winner. ## Quick Start Ask the agent to spike whether your idea is feasible, for example: "Spike whether I can stream LLM tokens over websockets to the browser with low latency before I commit to this architecture."