User shares trick to guide LLM agents during code generation

A Reddit user suggests inserting inline comments or instructions to steer an LLM agent's code output without restarting. This method prevents context loss from interruptions while still catching errors early.