cth9191

End-to-end conductor for cloning a piece of software. Walks the user through four gated phases — scope → deep research → plan (grilled + Codex-reviewed) → a paste-ready build prompt — composing the deep-research and grill-me-codex skills along the way. Use this whenever the user wants to clone, rebuild, reimplement, or make their own version of an existing app or tool ("clone WhisperFlow", "build my own Superhuman", "I want a local version of X", "reverse-engineer this app into a spec", "let's clone something"), or asks for a repeatable research→plan→build pipeline for replicating software. Trigger it even when the user just describes wanting to recreate an app's functionality without saying the word "clone". Invoke via /cloneify.

Interview-driven loop-engineering architect. Grills the user one question at a time about a task they want to automate, diagnoses where it sits on the maturity ladder (manual run → codified skill → automated → self-improving loop), forces the success-criteria decision through a 5-tier verification ladder, then BUILDS the complete runnable loop construct for their task right then: execution skill, trigger config, state store, verification block, stop rule, and README — with a validation gate baked in as the first run when the task is unproven. Then it offers a test run and, once they're happy, asks whether to wire it up to run on its own — cron, a Claude Code routine, an OS scheduled task, or left manual — and sets up that trigger so the loop is actually live and hands-off, not just a folder of files. Use this whenever the user wants to "loop engineer" something, build a self-improving / agentic loop, set up a recurring AI task that gets better over time, turn a manual workflow into an automated loop, build a