How does this skill prevent infinite loops and high costs?

The skill monitors for loop churn and cost drift. It provides recovery mechanisms to freeze the loop, perform an audit, and reduce the scope to specific failing units if measurable progress isn't being made.

What happens if the agent gets stuck during a task?

The skill provides recovery patterns such as the '/harness-audit' command, allowing users to replay failed segments with explicit acceptance criteria or manually reduce the scope to fix the root cause.

Can this skill handle pull request workflows?

Yes, it includes a specific 'continuous-pr' pattern designed for strict CI/PR control, ensuring that autonomous changes meet existing repository standards before merging.

What is a continuous agent loop in Claude Code?

A continuous agent loop is a structured workflow pattern that allows an AI agent to perform tasks autonomously by iterating through cycles of generation, testing, and refinement until a quality gate is passed.

Continuous Agent Loop

Name: Continuous Agent Loop
Author: saar120

bysaar120

0•

生産性とワークフロー

Orchestrates autonomous AI agent workflows with integrated quality gates, evaluation loops, and robust error recovery mechanisms.

The Continuous Agent Loop skill provides a sophisticated framework for managing long-running AI agent tasks within Claude Code. It implements a canonical selection logic that automatically determines the best execution pattern—whether that requires strict PR control, RFC-based decomposition, or parallel generation—to ensure high-quality outputs. By integrating session persistence and evaluation harnesses, it effectively mitigates common AI agent issues like loop churn, cost drift, and merge stalls, offering a production-grade environment for truly autonomous development.

主な機能

01Session persistence for long-running autonomous workflows

02Automated recovery controls for stalled or drifting agents

030 GitHub stars

04Intelligent loop selection flow for PR, RFC, and parallel generation

05Integrated quality gates and evaluation loop harnesses

06Cost-conscious escalation and churn prevention

ユースケース

01Executing autonomous PR review and fix cycles with strict CI/CD quality controls

02Running exploratory parallel code generation tasks for complex architectural problems

03Managing complex feature implementations through automated RFC-to-code pipelines

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry continuous-agent-loop

For use in Claude.ai and ChatGPT

主な機能

01Session persistence for long-running autonomous workflows

02Automated recovery controls for stalled or drifting agents

030 GitHub stars

04Intelligent loop selection flow for PR, RFC, and parallel generation

05Integrated quality gates and evaluation loop harnesses

06Cost-conscious escalation and churn prevention

ユースケース

01Executing autonomous PR review and fix cycles with strict CI/CD quality controls

02Running exploratory parallel code generation tasks for complex architectural problems

03Managing complex feature implementations through automated RFC-to-code pipelines

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry continuous-agent-loop

For use in Claude.ai and ChatGPT