I wanted to see how close I could get Claude to ~one-shot a typical startup while allowing for high-signal user input. I ended up with two Claude commands that handle the heavy lifting. Optimized for token efficiency, time (parallelization), surviving context flushes / session restarts.