Claude Code slash commands fail, show facts instead

This behavior points to a frustrating regression where the model interprets an action command as a knowledge query.

• –Failing to distinguish between deterministic commands and open-ended queries is a core failure mode for AI CLIs
• –Returning definitions instead of executing actions breaks user trust and disrupts automation workflows
• –The issue highlights the need for strict, non-LLM parsing for explicit slash commands
• –Anthropic has not yet publicly addressed whether this is a bug or a misguided feature update