What is a Continuous Agent Loop?

It is a structured execution pattern that allows AI agents to work autonomously through repetitive or complex tasks while maintaining strict quality standards and automated recovery protocols.

Can I use this for pull request management?

Yes, the skill includes a specific 'continuous-pr' selection flow designed for environments that require strict CI/PR control and quality gate integration.

How does this skill prevent infinite loops and cost drift?

The skill includes failure mode detection for loop churn and provides recovery commands such as freezing the loop, running audits, and reducing task scope to specific failing units.

What happens if an autonomous loop fails?

Recovery involves freezing the current loop, executing a harness audit, reducing the scope to the specific failing component, and replaying the task with explicit acceptance criteria.

Does it support parallel task execution?

Yes, the 'infinite' loop selection option is specifically designed for exploratory parallel generation when a broad search for solutions is required.

Continuous Agent Loop

Name: Continuous Agent Loop
Author: affaan-m

byaffaan-m

•

137,499

•

Productivity & Workflow

Implements robust, autonomous execution loops with built-in quality gates and recovery mechanisms for reliable agentic workflows.

The Continuous Agent Loop skill provides a standardized framework for managing autonomous AI agent behavior within Claude Code and similar environments. It offers a variety of specialized loop strategies—including RFC-based decomposition, CI/PR-controlled flows, and parallel exploratory generation—while ensuring reliability through integrated quality gates and evaluation harnesses. By providing explicit recovery protocols and failure mode detection, this skill prevents common agentic issues like infinite churn and cost drift, making it an essential component for production-grade autonomous software development.

Key Features

01Audit-ready evaluation loops to ensure measurable progress and accuracy

02Session persistence and state management for complex, multi-step tasks

03Integrated quality gates for enforcing code standards and performance metrics

04Dynamic loop selection flow including RFC-DAG and Continuous-PR strategies

05Automated recovery protocols to handle loop churn and merge queue stalls

06137,499 GitHub stars

Use Cases

01Managing long-running autonomous development tasks with strict quality controls

02Preventing cost escalation and resource drift during exploratory parallel code generation

03Implementing complex architectural changes that require RFC decomposition before execution

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add affaan-m/everything-claude-code continuous-agent-loop

For use in Claude.ai and ChatGPT

Key Features

01Audit-ready evaluation loops to ensure measurable progress and accuracy

02Session persistence and state management for complex, multi-step tasks

03Integrated quality gates for enforcing code standards and performance metrics

04Dynamic loop selection flow including RFC-DAG and Continuous-PR strategies

05Automated recovery protocols to handle loop churn and merge queue stalls

06137,499 GitHub stars

Use Cases

01Managing long-running autonomous development tasks with strict quality controls

02Preventing cost escalation and resource drift during exploratory parallel code generation

03Implementing complex architectural changes that require RFC decomposition before execution

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add affaan-m/everything-claude-code continuous-agent-loop

For use in Claude.ai and ChatGPT