My agent.md to improve LLM-assisted code quality

Guess which of these LLM outputs is watermarked

Vomit: Clean up Claude 5's token output with a separate LLM

A simple fix for LLM tail latency