Overview

Refactor a CLAUDE.md. Every line it keeps is loaded into every session the file covers, and a frontier model follows only around 150 to 200 standing instructions consistently, so each line competes for that budget. Everything else moves down the ladder or out of the file.

The file serves every model the user runs, not only the one this session runs on, so audit it against every model the user runs. Claude Code assembles its system prompt per model: a line that merely restates the prompt under one model can be the only copy under another.

Instructions

1. Pick the target. If the invocation names one, use it. Otherwise ask with AskUserQuestion, one option per file that exists, with the resolved path in the label:

   • The user-level CLAUDE.md at ~/.claude/CLAUDE.md.
   • The current project's CLAUDE.md.

   Then settle the models to audit: the ones the invocation names, otherwise ask in the same question call, multi-select, with this session's model as the first option. Resolve each to its full model ID from the model list in your system prompt.

2. Fetch the guides. Fetch these pages. They calibrate the delete and rewrite verdicts below:

   • https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices
   • The page for each audited model, regardless of which one this session runs on, plus any earlier model's page it names as its baseline: https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-<model-id>