autonomous-loops

Orchestrate iterative Claude-based development loops with CI checks and retries.

1|Updated Mar 18, 2026
One-click install
npx skills add https://github.com/xxih/ai-harness-zh --skill autonomous-loops-xxih
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: autonomous-loops
Source: https://github.com/xxih/ai-harness-zh/tree/main/references/translations/everything-claude-code/docs/zh-CN/skills/autonomous-loops
Command: npx skills add https://github.com/xxih/ai-harness-zh --skill autonomous-loops-xxih

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Enables reliable automation and orchestration of iterative Claude-based development workflows by bridging per-iteration context, managing parallel agent runs, and handling CI/merge lifecycle concerns to avoid endless, unsafe, or uncoordinated loops.

Core Features & Use Cases

  • Sequential pipelines: script a series of isolated non-interactive Claude steps for TDD, cleanup, verification, and commit.
  • Persistent REPL sessions: run NanoClaw-style sessions that preserve conversation history and incremental context across restarts.
  • Continuous PR loops: create, test, repair, and merge PRs in an automated loop with CI monitoring and retry logic.
  • Multi-agent DAG orchestration: decompose RFCs into work units, run layered quality pipelines, handle merge-queue eviction, and recover from conflicts with contextual guidance.
  • Quality and safety gates: insert de-sloppify cleanup passes, CI-driven fixes, and completion signals to bound cost and runtime.

Quick Start

Use the autonomous-loops skill to run a continuous Claude iteration that creates a feature branch, runs tests, opens a pull request, and retries fixes until CI passes or the configured completion signal stops the loop.

Frequently Asked Questions about autonomous-loops

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

FAQPage Schema
How do I automate continuous PR loops with Claude for iterative code generation?

You can automate continuous PR loops by orchestrating Claude to create feature branches, run tests, open pull requests, and retry fixes until CI passes or a configured completion signal stops the loop.

How does multi-agent DAG orchestration work for decomposing RFCs into development tasks?

Multi-agent DAG orchestration works by decomposing RFCs into work units, running layered quality pipelines, and handling merge-queue eviction with contextual guidance to coordinate parallel agent runs.

Can I run persistent NanoClaw REPL sessions that preserve conversation history across restarts?

Yes, you can run NanoClaw-style persistent REPL sessions that preserve conversation history and incremental context across restarts to maintain state during interactive development.

What is the best way to bound cost and runtime for autonomous Claude development loops?

The best way to bound cost and runtime is to configure exit conditions like max runs, cost limits, or completion signals, and insert de-sloppify cleanup passes and CI-driven fixes.

How do I handle merge-queue eviction and conflicts in automated CI-driven PR iterations?

You handle merge-queue eviction and conflicts by using contextual guidance to recover from issues, applying automated CI check and retry flows within the orchestration loop.