Autonomous Loops

Documents patterns and architectures for running Claude Code autonomously in loops, from simple pipelines to multi-agent DAG orchestration.

What it does

This skill is a reference guide describing loop patterns for running Claude Code non-interactively, ranging from simple sequential claude -p pipelines to RFC-driven multi-agent DAG orchestration. It presents a spectrum table of patterns with complexity levels and use cases, and illustrates the simplest pattern with a shell pipeline example covering implementation, cleanup, verification, and commit steps. It also lists design principles such as step isolation, ordering, avoiding negative instructions, and exit-code propagation. It produces guidance and reference material rather than executable artifacts.

When to use it

Use it when setting up automated development workflows that run without human intervention, or when choosing a loop architecture for a given problem. It also fits building CI/CD-style continuous development pipelines, running parallel agents with merge coordination, or adding quality gates and cleanup passes to automated workflows.

Requirements

No scripts or assets are included; it is instructions and reference documentation only. Following the examples assumes access to Claude Code and its non-interactive claude -p mode, plus a shell environment for the pipeline examples.