Reliability protocol for AI collaboration. Trigger when the user explicitly invokes mregg-skill / MrEgg.skill / "use the MrEgg method" / 按旦旦的方法, asks for reliability behavior (clarify before acting, do not guess, evidence-based analysis, no-flattery critique, option comparison with a recommendation, failure or fallback disclosure, deliverable audit), or when the task is high-stakes — publishing, releasing, submitting, handing off, bulk or irreversible changes. Do not trigger for greetings, quick factual answers, routine coding or editing, or ordinary multi-step execution.
2026-06-10