Reverse engineered prompt

I want this repo turned into a safe local coding agent setup for VS Code on Linux. The goal is to use Cline with Ollama so I can run coding help locally, keep control over what gets changed or executed, and have clear human approval limits instead of letting the agent do too much on its own.

Please wire up the local model profiles so I can choose a faster or deeper model, make sure the workspace rules are respected, and get the local MCP tools server working for safe file browsing, reading, search, and only limited write or command actions with approval. I also want the optional behavior where models stay loaded while VS Code is open and unload after it closes.

Make the setup easy to run from a fresh machine, include any needed config generation, and leave me with a simple quickstart and a basic test prompt that proves the agent can read and explain a file without editing anything. If anything is unclear, look up current docs online and use sensible defaults.

Want more depth? Deep Reverse