What problem does it solve? Engineers often hit gnarly topics — a recurring bug family, a subtle design constraint, an unfamiliar system — that no one has time to explain properly. This Skill turns one of those topics into a verified, layered explainer doc and then walks you through it interactively until you genuinely understand it, ending in a decision rather than a lecture. ## Core Features & Use Cases - Isolated investigation and authoring: Delegates research to an isolated subagent so file reads and repro attempts never clutter your session, producing a verified explainer doc with claims marked as verified versus inferred. - Interactive layered walkthrough: Runs an existing explainer doc one layer per turn with an explicit checkpoint after each layer, so you control the pace and can ask questions in place. - Live challenge handling: When you push back on a claim, it designs and runs a discriminating experiment live, correcting the doc if the evidence contradicts it. - Use Case: After a confusing gate or CI finding, run the author mode on that bug family to produce a durable explainer doc, then walk it layer by layer in a fresh session before merging. ## Quick Start Ask the assistant to explain the topic behind your latest CI failure by invoking the explain skill with that topic as the argument.