What problem does it solve? It prevents AI agents and developers from making changes to the VuaAssistant repository that contradict the product vision, skip real verification, or break working subsystems like the AI Router sidecar and agent-runner. ## Core Features & Use Cases - Product-idea alignment guard: Requires reading idea.md before proposing changes, with hard rules such as subscription-first OAuth login, no Bash/Terminal exposure to models, and an internal AES-256 Vault. - Three-level verification workflow: Mandates compile checks, tests, and real hands-on testing of the macOS app before reporting any task as done. - Sidecar and release discipline: Covers agent-runner TypeScript builds, better-sqlite3 native module version matching, background process supervision with restart caps, and PO-controlled version tagging. - Use Case: Before refactoring the sign-in screen, an agent loads this skill, confirms the current OAuth-first connection order is intentional per idea.md, and avoids a costly incorrect rewrite. ## Quick Start Load the vuaassistant-dev-guidelines skill before modifying anything in the vuaassistant repository and follow its verification and release rules.