loop-engineering

What problem does it solve?

This Skill turns repeated agent work into a controlled loop so tasks do not drift into endless retries, self-grading, or unverified claims. It helps you decide when a task should be a one-shot action, a goal-driven loop, an automation, or an auto-research workflow.

Core Features & Use Cases

• Loop Design: Define the trigger, objective, scope, evidence, limits, and permission boundary before any repeated execution starts.
• Topology Selection: Choose the simplest workable pattern, such as single-agent, maker-checker, or manager-workers, based on whether independent verification is needed.
• Failure Prevention: Detect stalled loops, runaway iteration, vague stop conditions, and unsupported writes before they waste time or mutate shared state.
• Use Case: A CI repair task can be converted into a bounded maker-checker loop with explicit evidence, a finite budget, and a clear recovery path.

Quick Start

Use loop-engineering to define the loop mode, trigger, scope, evidence, limits, and verifier for this task.