What problem does it solve? Starting a new project with an AI assistant often means re-explaining context every session. This Skill captures the project context once by writing a CLAUDE.md contract through structured closed-ended questions, so future sessions never need to re-ask what is already decided. ## Core Features & Use Cases - Closed-Question Framing: Asks only yes/no or multiple-choice questions (max 4 per round) with a recommended option, covering business context, scope, stack, and working method. - CLAUDE.md Generation: Writes a concise, validated CLAUDE.md at the project root following a strict template, marking undecided items as _à décider_ instead of inventing values. - BRAIN & Error Journal Bootstrap: Creates ~/brain/<PROJECT_KEY>/ with rules/, an empty bag.ndjson, and appends an error-logging block to CLAUDE.md in append-only mode. - Use Case: When starting a new client project, trigger this Skill to produce a validated CLAUDE.md and an initialized error journal before any code is written. ## Quick Start Ask the assistant to initialize the project by saying "init projet" and answer the closed-ended questions to generate the CLAUDE.md.
