Your CLAUDE.md might be sabotaging $OPUS 5.5 performance.
Old prompting habits like "think step by step" or "double-check your work" can actually slow down newer models instead of helping them.
Anthropic's Lance Martin just dropped a quick fix:
Open Claude Code and run:
/claude-api prompt-audit
It scans your skills, CLAUDE.md, and prompts—strips out outdated instructions and rewrites them based on the latest Opus 5.5 guidance.
Takes a few minutes. Worth running tonight if you're serious about optimizing your workflow.
Old prompting habits like "think step by step" or "double-check your work" can actually slow down newer models instead of helping them.
Anthropic's Lance Martin just dropped a quick fix:
Open Claude Code and run:
/claude-api prompt-audit
It scans your skills, CLAUDE.md, and prompts—strips out outdated instructions and rewrites them based on the latest Opus 5.5 guidance.
Takes a few minutes. Worth running tonight if you're serious about optimizing your workflow.