autonomous-loops

Orchestrate autonomous Claude Code loops for multi-agent pipelines and continuous iterations.

1|Updated Mar 5, 2026
One-click install
npx skills add https://github.com/beratcelik1/compound-claude --skill autonomous-loops-beratcelik1
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: autonomous-loops
Source: https://github.com/beratcelik1/compound-claude/tree/main/skills/autonomous-loops
Command: npx skills add https://github.com/beratcelik1/compound-claude --skill autonomous-loops-beratcelik1

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill provides patterns, architectures, and operational guidance to run reliable autonomous Claude Code loops and multi-agent pipelines, preventing context drift, merge conflicts, and repeated CI failures across iterations.

Core Features & Use Cases

  • Sequential Pipeline: Run isolated, ordered claude -p steps for deterministic CI-style workflows and small feature changes.
  • NanoClaw REPL & Persistent Sessions: Maintain session history and continuity for interactive exploration and incremental work.
  • Infinite Agentic Loop & Parallel Agents: Orchestrate spec-driven parallel generation with assigned creative directions to avoid duplication.
  • Continuous Claude PR Loop: Create, test, auto-fix, and merge PRs with CI gate handling and shared cross-iteration notes.
  • Ralphinho / RFC-Driven DAG: Decompose RFCs into dependency DAGs, run tiered quality pipelines per unit, and use a merge queue with eviction recovery.
  • De-Sloppify Pattern: Add focused cleanup passes after implementation to remove test and code slop without constraining authors.

Quick Start

Create a sequential autonomous loop that implements the feature in docs/auth-spec.md, runs a de-sloppify cleanup pass, verifies tests and CI, and persists progress to SHARED_TASK_NOTES.md.

Frequently Asked Questions about autonomous-loops

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I orchestrate autonomous Claude Code loops for continuous PR integration?

You can orchestrate autonomous Claude Code loops by using a continuous PR loop pattern that creates, tests, auto-fixes, and merges pull requests while handling CI gate recovery and persisting shared cross-iteration notes.

What is the best way to run multi-agent pipelines without context drift?

To prevent context drift in multi-agent pipelines, use isolated context windows for each agent and maintain persistent cross-iteration state, ensuring parallel generation follows assigned creative directions without duplication.

How does RFC-driven DAG orchestration work for parallel development units?

RFC-driven DAG orchestration decomposes RFCs into dependency DAGs, runs tiered quality pipelines per unit, and uses a merge queue with eviction recovery to manage multi-unit parallel work.

Can I run sequential CI-style workflows with Claude Code?

Yes, you can run isolated, ordered sequential pipeline steps for deterministic CI-style workflows, which are suitable for small feature changes and structured repository-level tasks.

Why do I need a de-sloppify pass after autonomous code generation?

A de-sloppify pass is needed after autonomous code generation to add focused cleanup that removes test and code slop without constraining authors, ensuring tiered quality pipeline standards are met.

How do I handle merge queue eviction during autonomous development loops?

You handle merge queue eviction by implementing recovery patterns within the DAG orchestration, allowing the autonomous loop to resolve conflicts and re-queue units without losing persistent cross-iteration state.