You are an expert in prompt engineering for Claude, specializing in optimizing CLAUDE.md instruction files. Your task is to analyze the current session and evolve the CLAUDE.md to prevent recurring problems.
This prompt is PROJECT-AGNOSTIC. Use relative paths and dynamic discovery.
Before analyzing CLAUDE.md, mine this conversation for evidence of friction.
There's a discussion here in Reddit about using Claude to incrementally improve CLAUDE.md based on the current context.
I had Claude argue with itself in order to improve the prompt I got from there. This follows Andrej Karpathy's LLM Council pattern. I find that having Claude argue with Gemini works best (e.g., with the gemini-collab MCP server), but here I just let it argue with itself.
You can either grab the improved prompt in self-improve.md, or redo this yourself. To redo it, first grab a-c-m's tweak from https://gist.github.com/a-c-m/f4cead5ca125d2eaad073dfd71efbcfc and put it in ~/.claude/commands/self-improve.md. Then, run the following prompt in Claude Code:
Simulate two experts in Anthropic prompt engineering, slash commands, agents, skills, etc. discussing the prompt here: ~/.claude/commands/self-improve.md