The prompt engineer for autonomous agent runs. Interviews the user about what they're building, then writes the perfect mission prompt — never-ask autonomy rules, multi-agent orchestration, end-to-end UI testing quality bar, and a provable definition of done baked in — plus a full overnight scaffold (feature list, progress state, launcher) when the run is big. Works for Claude Code, Codex, Cursor, or any agent. Use when the user invokes /loop-maker, says "loop", "make me a loop", "write me a prompt", "help me plan this feature", "make the perfect prompt", "run this overnight", "ship while I sleep", "goal prompt", "master prompt", "autonomous run", "work all night", "build me a company", or is planning features they want an agent to build and test without stopping.
2026-07-09