model-prompt-adapter

Universal prompt addenda that make workspace files work across different LLM providers when using fallback chains. Addresses documented failure modes per model family (scope creep, prompt leaking, over-eagerness, fabricated completion). Use when: (1) running multi-model fallback (e.g., Claude primary → GPT fallback), (2) optimizing injected workspace files for cross-model compatibility, (3) mitigating known provider-specific failure modes in agent systems.