Does this skill work with existing Claude Code hooks?

Yes, it is designed to work alongside other hooks like Prettier or linting tools, and can be integrated with a hooks-manager for coordinated execution.

How does the skill prevent runaway costs?

The skill includes configurable safety limits for maximum iterations, total USD spend, and maximum runtime, automatically stopping the process once any threshold is reached.

What is the main benefit of the Autonomous Opus Loop?

It allows Claude Code to work on long, multi-step tasks without requiring human intervention between every step, using a second AI model to ensure progress remains on track toward your goals.

How do I stop an autonomous session early?

You can manually interrupt the process with Ctrl+C or use the provided emergency-stop.sh script to immediately terminate the loop and reset the hook status.

Can I use models other than Opus 4.5 for the analyzer?

Yes, the configuration allows you to specify the analyzer model, such as Claude 3.5 Sonnet, via the analyzer_model setting in your autonomous-config.json file.

Autonomous Opus Loop

Name: Autonomous Opus Loop
Author: adaptationio

byadaptationio

生産性とワークフロー

Transforms Claude Code into a fully autonomous agent by utilizing a secondary model instance to evaluate progress and direct continuous development workflows.

概要

This skill elevates Claude Code from an interactive tool to a self-directing development agent by establishing a collaborative feedback loop between two Claude instances. Using a Stop hook architecture, an external Opus 4.5 analyzer reviews work against your defined success criteria after every iteration, deciding whether to continue with specific next-step instructions or terminate upon completion. It is specifically designed for complex, multi-step engineering tasks—such as overnight feature builds or deep codebase refactoring—while maintaining safety through robust cost tracking, iteration limits, and automated doom-loop prevention.

主な機能

Configurable success criteria and objective-driven task management
0 GitHub stars
Automatic doom-loop and circular work detection via semantic analysis
Integrated safety controls for cost, iterations, and runtime limits
Detailed session logging and real-time status tracking via CLI
Dual-model collaboration for objective progress analysis and direction

ユースケース

Comprehensive codebase refactoring to meet strict architectural standards
Autonomous overnight feature development and large-scale implementation
Deep-dive bug hunting and memory leak investigation sessions

概要

主な機能

Configurable success criteria and objective-driven task management
0 GitHub stars
Automatic doom-loop and circular work detection via semantic analysis
Integrated safety controls for cost, iterations, and runtime limits
Detailed session logging and real-time status tracking via CLI
Dual-model collaboration for objective progress analysis and direction

ユースケース

Comprehensive codebase refactoring to meet strict architectural standards
Autonomous overnight feature development and large-scale implementation
Deep-dive bug hunting and memory leak investigation sessions