What problem does it solve? Working on Lovable.dev repositories with coding agents requires knowing the project architecture (Vite SPA vs TanStack Start), where configuration lives, and how to keep secrets out of generated instructions. This Skill provides the shared project context and rules so agents like Codex and Claude Code operate on a Lovable repo correctly from the first step. ## Core Features & Use Cases - Architecture Detection: Distinguishes Vite SPA (CSR) from TanStack Start (SSR) projects using app.config.ts, vite.config.ts, and the detect-architecture.py helper script. - Provider-Neutral Configuration: Reads .lovable-agent/config.json and context.md as the source of truth while preserving CLAUDE.md and .claude/ as compatible adapters. - Secret-Safe Context Handling: Records secret names and status only, never values, preview tokens, or API keys in config, logs, or generated reports. - Use Case: An agent opens a Lovable repository, reads .lovable-agent/config.json, detects a TanStack Start architecture, and generates correct instruction shims without leaking credentials. ## Quick Start Use the lovable-project skill to inspect this Lovable repository, detect whether it is a Vite SPA or TanStack Start app, and summarize its agent configuration.