Claude Code cut over 80% of its system prompt, scores held
Over 80% of Claude Code's system prompt is gone for Claude Opus 5 and Claude Fable 5, and Anthropic's coding evaluations show no measurable loss.
Old lines like "default to writing no comments" clashed with skills and user requests, so Claude had to untangle conflicting orders. The new prompt asks it to match the surrounding code's comment density, naming, and idiom.
Much of the rest moved to progressive disclosure. Code review and verification became skills Claude calls when needed, and some tools load full definitions only through ToolSearch, like a manual where you open just the chapter you need.
To trim your own setup, /doctor in Claude Code rightsizes skills and CLAUDE.md files. In our view, the 80% figure would carry more weight if Anthropic had named the evaluations behind it.
Related stories
- Three habits could be holding Opus 5.5 and Sonnet 5.5 back
- Delete "think carefully" from your Opus 5.5 prompts
- "Verify twice" wastes tokens on Opus 5, Anthropic says
- A hidden system prompt paragraph tells Claude Code to stop asking permission
- Anthropic cut roughly 80% of Claude Code's system prompt
- Anthropic rewrites its context engineering rules for Claude 5
Comments
No comments yet. Be the first.
Join the conversation
Sign in with Google to leave a comment. Your name and avatar come from your Google profile, and the comment appears after moderation.