cc-ralph

Automate receipt-verified task execution loops across multiple Claude sessions.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/z23cc/cc-code --skill cc-ralph
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: cc-ralph
Source: https://github.com/z23cc/cc-code/tree/main/skills/cc-ralph
Command: npx skills add https://github.com/z23cc/cc-code --skill cc-ralph

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires cc-work, cc-teams, and includes scripts (resource) components.

What problem does it solve?

Ralph provides an autonomous execution loop with multi-model review gates and receipt-based verification, spawning fresh Claude sessions per iteration to avoid context bleed and ensure verifiable progress.

Core Features & Use Cases

  • Fresh Claude sessions per iteration to prevent context pollution
  • Multi-model review gates for plan, work, and completion verification
  • Receipt-based proof-of-work to validate task completion
  • Safety guards, sentinel controls, and configurability for unattended operation
  • End-to-end orchestration for epics and iterative tasks across multiple teams

Quick Start

Initialize Ralph with /cc-ralph-init and start the main loop using the ralph.sh script.

Frequently Asked Questions about cc-ralph

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

FAQPage Schema
How do I automate unattended task execution across multiple Claude sessions?

Automate unattended task execution across multiple Claude sessions by running an autonomous execution loop that spawns fresh contexts per iteration, preventing context bleed. This approach uses receipt-based verification to validate progress and ensure traceable task completion without manual intervention.

Why does context pollution happen during iterative workflow automation and how can I prevent it?

Context pollution occurs when a single session accumulates irrelevant state over long iterative workflows. Prevent it by spawning fresh Claude sessions per iteration, ensuring clean execution contexts for each task cycle while maintaining verifiable progress through receipt-backed proof-of-work.

Can I use multi-model review gates to verify plan and completion stages in autonomous workflows?

Yes, multi-model review gates can verify plan, work, and completion stages in autonomous workflows. These review gates act as safety hooks within the execution loop, validating task quality and ensuring risk-guarded automation before proceeding to the next iteration.

Do I need cc-work and cc-teams to run unattended automation loops with safety guards?

Yes, cc-work and cc-teams are required dependencies to run unattended automation loops with safety guards. These provide the foundational orchestration and multi-team coordination needed for end-to-end epic execution and configurable loop controls.

What is the best way to validate task completion in an autonomous execution loop?

Validate task completion in an autonomous execution loop using receipt-based proof-of-work. Receipts validate that each iteration has produced verifiable progress, ensuring risk-guarded automation with safety hooks and sentinels has successfully finished before advancing.

How do I initialize an autonomous execution loop with configurable sentinels?

Initialize an autonomous execution loop with configurable sentinels by running the ralph.sh script after setup. This starts the main loop, applying safety guards and sentinel controls for unattended operation while managing end-to-end epic execution across multiple sessions.