External LLM Consulting

Audience: Developers seeking alternative AI perspectives on architecture and design decisions. Goal: Invoke external LLM CLIs in read-only/sandbox mode to get structured second opinions, then present results for comparison with Claude's perspective.

When to Use

• Validating architectural choices with diverse AI perspectives
• Exploring alternative approaches Claude might not consider
• High-stakes design decisions benefiting from model diversity
• Surfacing blind spots through cross-model analysis

General Workflow

1. Gather relevant context (CLAUDE.md, AGENTS.md, existing code patterns, constraints)
2. Formulate a focused prompt with context, specific questions, and structured output request
3. Execute the external CLI in sandbox/read-only mode
4. Parse and structure the response for comparison