What problem does it solve? New AI sessions on an undocumented repository start with zero context, forcing developers to re-explain the tech stack, conventions, and commands every time. This Skill scans the codebase once and generates persistent project context files so every future session starts informed. ## Core Features & Use Cases - Automated Tech Stack Detection: Scans config files (package.json, pyproject.toml, Cargo.toml, go.mod) to identify language, framework, package manager, test framework, and linter without guessing. - Context File Generation: Creates CLAUDE.md plus a .rune/ directory containing conventions, decisions, progress, session-log, instincts, contract, INVARIANTS.md, and a human-readable DEVELOPER-GUIDE.md. - Invariant Seeding: Runs bundled Node.js scripts to detect danger zones, shared constants, state machines, and mirrored literal tuples, writing them to .rune/INVARIANTS.md with merge-safe re-runs. - Use Case: A developer clones an unfamiliar legacy repository with no documentation. Running the onboard flow produces a complete CLAUDE.md with verified build/test commands and a developer guide, so both AI sessions and new team members start with full context. ## Quick Start Ask the AI to run the rune onboard flow on this repository to scan the codebase and generate CLAUDE.md and the .rune context directory.