Ralph CI is a specialized infrastructure skill that adds automation, safety guardrails, and reporting to autonomous AI development loops orchestrated by Claude Code.

How does the budget control work?

The harness scripts track API usage costs during each iteration and will automatically terminate the loop if the total spend exceeds your pre-configured USD limit.

Does Ralph CI require specific project files?

Yes, it is a downstream consumer of the 'add-project' skill and requires an existing .ralph/ directory containing PLAN.md and .ralphrc files to function correctly.

What happens if an agent gets stuck in a loop?

Ralph CI includes a circuit breaker that monitors for 'no progress' or 'repeated errors.' If the agent fails to advance the project plan after a set number of attempts, the system triggers a safety shutdown.

Can I get notified if a CI loop fails?

Yes, Ralph CI supports real-time notifications via Webhooks (for Slack/Discord) and SMS alerts (via n8n) for events like success, failure, budget limits, or emergency kills.

Ralph CI: Reliable Loop Infrastructure

Name: Ralph CI: Reliable Loop Infrastructure
Author: mattwoodco

bymattwoodco

•

Deployment & DevOps

Automates autonomous agent loops with deterministic success gates, cost controls, and safety mechanisms for GitHub Actions or VPS environments.

Ralph CI provides a robust execution framework for autonomous development cycles, transforming standard Claude Code projects into reliable, managed agentic loops. It features two operational modes—GitHub Actions for matrix parallelism and VPS (PM2) for persistent execution—while enforcing strict guardrails like budget caps, circuit breakers for stalled progress, and emergency kill switches. Designed as a downstream enhancement for Ralph-enabled projects, it handles the orchestration, reporting, and safety logic required to run AI agents at scale with confidence.

Key Features

01Deterministic success gates that validate builds, typechecks, linting, and tests

023 GitHub stars

03Automated cost controls and budget caps to prevent unexpected API spend

04Integrated reporting via Webhooks and SMS for real-time status updates on agent progress

05Dual-mode execution support for GitHub Actions matrix runners or VPS (PM2) environments

06Advanced safety mechanisms including circuit breakers, zombie process protection, and emergency kill switches

Use Cases

01Implementing persistent, long-running agent tasks on a dedicated VPS with PM2 process management

02Scaling autonomous code generation across multiple parallel GitHub Action runners

03Enforcing strict quality gates and budget limits on AI-driven development workflows

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mattwoodco/skills ralph-ci

For use in Claude.ai and ChatGPT

Download Skill