What problem does it solve? Raw feature ideas arrive as unstructured one-liners with no owner, no success criteria, and no traceability. This Skill turns a single sentence into a structured intake document with a unique FEAT-ID, explicit open questions, and an AI-classification flag, so the downstream PRD phase has a concrete artifact to start from. ## Core Features & Use Cases - Structured intake: Writes a feature-request document from a template, filling what can be inferred and marking everything else as answerable TBD questions instead of inventing content. - FEAT-ID allocation and state tracking: Assigns the next free zero-padded ID across product docs, state, and archive, then opens a state machine entry with phase, branch name, and content hash. - AI surface classification: Tags the feature as none, llm-generation, llm-extraction, llm-classification, rag, or agentic, which determines whether eval gates apply downstream. - Use Case: A product owner types "/feature add semantic search to the help center" and receives a draft intake document with open questions about the success metric and constraints, plus a G0 gate verdict listing exactly what must be answered before the PRD phase. ## Quick Start Invoke /feature followed by a one-line feature request in quotes to generate the intake document and state entry.
