experimental-research-metacognition

What problem does it solve? When documentation conflicts with actual behavior, bugs are suspected, or implementation choices are contested, guessing wastes time. This Skill provides a disciplined workflow for answering technical questions through falsifiable experiments instead of assumptions. ## Core Features & Use Cases - Structured Research Loop: Define a research question, falsifiable hypothesis, and success criteria before writing any code, then iterate through a tight run-observe-update cycle. - Minimal Reproduction Guidance: Choose the smallest experiment type that can falsify a hypothesis, from pure unit tests and check scripts to contract tests, Puppeteer runs, and performance benchmarks. - Metacognition Guardrails: Apply stop rules to avoid rabbit holes, track confidence scores from 0 (guess) to 3 (regression-guarded), and distill findings into durable documentation, Skills, or Patterns. - Use Case: When unsure whether a server component renders correctly under SSR, create a session, write a hypothesis, build a small check script under checks/, run it, and record the evidence before committing to an implementation approach. ## Quick Start Ask the AI to investigate an uncertain behavior by forming a hypothesis and building the smallest experiment that proves or disproves it.