Audit Reveals Claude Code Setup Violates Anthropic's Context Rules

A team audited their Claude Code configuration against Anthropic's context-engineering guidelines and found redundant context injection, violating rules like keeping CLAUDE.md lightweight and using progressive disclosure. They traced the issue to session-start hooks that loaded overlapping data, increasing token waste. The fix involved removing duplicate commands and trusting the model's memory system.