Anthropic Releases Automated Prompt Audit Tool to Eliminate Legacy Instructions from Claude Code

Anthropic engineer Lance Martin has released /claude-api prompt-audit, a Claude Code command that automatically scans and fixes outdated prompting patterns in skills, documentation, and configuration files. The tool addresses a widespread issue: as Claude models grow more capable, instructions written for earlier versions often degrade performance by creating unnecessary verification steps, excessive emphasis language, and rigid scaffolding that newer models follow too literally.