What problem does it solve? After an AI agent answers a question by drilling into source code, the discovered facts are lost unless someone manually documents them. This Skill captures those verified findings from the conversation and writes them into the project's .Knowledge base so future sessions can route to them directly. ## Core Features & Use Cases - Automatic knowledge distillation: Extracts reusable facts (mechanisms, state transitions, error contracts, config effects) from the latest question-and-answer round, with source file and line references. - Depth-aware merge decisions: Scores drill-down depth and description granularity to decide whether to append to an existing topic, create a sub-topic, or create a new module topic with a stock-doc. - Routing synchronization: Updates manifest-routing.json, matcher includeAny keywords, taskToTopicRules, and index.md so new topics are discoverable, then runs flow2spec kb plan/apply/build/check automatically. - Use Case: After asking how a retry mechanism works and getting a source-backed answer, trigger the distill flow to persist those findings as a knowledge topic that the next session will hit via keyword routing. ## Quick Start Ask the agent to distill the knowledge from our last Q&A exchange into the knowledge base using f2s-kb-distill.
