claude-loop

Orchestrate sequential autonomous coding workflows from PRD stories via fresh subagents per Task.

48|7|Updated May 16, 2026

Search Tags:#orchestration #subagent #claude-code #code-review #prd-json #autonomous-workflow #progress-log

Ready to run

npx skills add https://github.com/aronprins/claude-loop --skill claude-loop

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: claude-loop
Source: https://github.com/aronprins/claude-loop/tree/main
Command: npx skills add https://github.com/aronprins/claude-loop --skill claude-loop

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Manages autonomous, multi-story software work by orchestrating a fresh subagent for each PRD user story, keeping context separate and outputs verifiable.

Core Features & Use Cases

Fresh subagents per story to isolate work and reduce context leakage
Progress tracking via claude-loop/prd.json, claude-loop/progress.txt, and AGENTS.md-based knowledge sharing
Automatic verification of each subagent's output (typecheck, tests, and commits)
Branch management and archiving on completion or branch changes
Deterministic, sequential execution to avoid race conditions in git and PRD updates

Quick Start

Place your PRD at claude-loop/prd.json and say 'run the loop' to begin.

Frequently Asked Questions about claude-loop

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

FAQPage Schema

How do I orchestrate autonomous coding workflows for multiple user stories?▼

Autonomous coding workflows can be orchestrated by reading claude-loop/prd.json and spawning a fresh subagent per story via Task. This isolates work, reduces context leakage, and ensures typecheck and tests pass for each sequential story.

How do I prevent context leakage when running sequential LLM coding agents?▼

Preventing context leakage when running sequential LLM coding agents is achieved by spawning a fresh subagent per story. Context separation is maintained through git history, progress.txt, and AGENTS.md files for knowledge sharing between agents.

What's the best way to track progress for autonomous subagent code generation?▼

Tracking progress for autonomous subagent code generation is handled via claude-loop/prd.json and claude-loop/progress.txt files. Branch management and archiving occur on completion or branch changes to maintain deterministic, sequential execution.

Do I need a PRD file to start autonomous story-by-story code generation?▼

A PRD file is required to start autonomous story-by-story code generation. You must place your PRD at claude-loop/prd.json and instruct the agent to run the loop to begin sequential execution with strict per-story scope enforcement.

Why does my autonomous coding workflow report blockers during sequential story execution?▼

Blockers during sequential story execution are reported when a story cannot be completed. The workflow enforces strict per-story scope and requires minimal changes to the PRD until a story completes to avoid race conditions in git and PRD updates.