What problem does it solve? Choosing how to automate recurring or iterative agent work is confusing: should you use a plain prompt, a goal-based loop, a scheduled trigger, or a fully proactive routine? This Skill provides a decision framework that matches each task to the simplest loop primitive and ensures every loop has an explicit stop condition and usage bound. ## Core Features & Use Cases - Loop Selection Table: Routes tasks across four primitives — turn-based, goal-based (/goal), time-based (/loop, /schedule), and proactive compositions — based on what you hand off to the agent. - Stop Condition Design: Guides you to define deterministic completion criteria, turn caps, and token budgets so loops never run unbounded. - Quality and Cost Management: Covers verification skills, second-agent review, model routing, and usage inspection via /usage, /goal, and /workflows. - Use Case: You want an agent to keep fixing CI failures on your PR until it passes. The Skill routes you to a goal-based loop with a turn cap, e.g. "/goal get CI green, stop after 5 tries." ## Quick Start Ask the agent to help you pick the right loop primitive for your recurring task and define its stop condition and turn cap.
