Rather than adding an agents.md file written to stop LLM agents, I’m thinking about putting random comments into my code telling the agent to stop and warn the user, also some comments that are just word salads/non-sequiturs.
Rather than adding an agents.md file written to stop LLM agents, I’m thinking about putting random comments into my code telling the agent to stop and warn the user, also some comments that are just word salads/non-sequiturs.
If you can at all, I’d recommend not fucking up your codebase, but rather adding layers (reverse proxies or whatever) in front of your git repo hosting that will poison LLMs. So-called “tarpits” for LLMs. To whatever extent possible, don’t make your human users suffer for the sins and LLM scapers of tech bros.