OpenAI's Codex CLI ships a /goal command implementing autonomous looping — set an objective and the agent iterates until it evaluates completion or exhausts its token budget. Commands include /goal pause, /goal resume, and /goal clear, with states like pursuing, paused, achieved, unmet, and budget-limited. Simon Willison notes this is OpenAI's take on the 'Ralph loop,' implemented via auto-injected prompt templates (goals/continuation.md and goals/budget_limit.md) rather than hardcoded logic.