What problem does it solve? Engineering teams often make technology or architecture decisions on gut feel, or let exploratory work balloon into unplanned implementation. This Skill structures feasibility questions, technology comparisons, and proof-of-concept work into a time-boxed investigation whose only deliverable is a decision artifact, keeping throwaway spike code out of production. ## Core Features & Use Cases - Framed Investigation: Forces an explicit question, hard time budget, and decision criteria before any research begins, turning vague asks into testable questions. - Evidence-Gated Decision Artifact: Produces a findings document in .ai/ or an ADR in wiki/architecture/ with options, risks, and verifiable evidence, then runs a mandatory self-gate that rejects fabricated benchmarks, untested assumptions, and inflated confidence. - Clear Outcomes and Handoff: Closes every spike with one of three outcomes (decision made, blocked on info, or do-not-build) and routes architectural decisions into high-level design review before any build starts. - Use Case: Ask whether Kafka can handle 50k events/sec at p99 under 20ms on your infrastructure tier; the Skill benchmarks options, prototypes the riskiest assumption, and commits a decision artifact with cited evidence. ## Quick Start Invoke the research-spike skill with a specific feasibility question, a time budget, and the decision criteria you want evaluated.
