Verified:GitHub

ralph

Iterate tasks to verified completion with architect review and persistence loops.

Updated Feb 20, 2026

Search Tags:#parallel execution #agent orchestration #code completion #product requirements #task iteration #architect verification

Ready to run

npx skills add https://github.com/ssccio/cc-on-web-base --skill ralph-ssccio

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: ralph
Source: https://github.com/ssccio/cc-on-web-base/tree/main/skills/ralph
Command: npx skills add https://github.com/ssccio/cc-on-web-base --skill ralph-ssccio

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Ralph helps you keep working on a task until completion by enforcing a structured loop and requiring architect verification before declaring the work done.

Core Features & Use Cases

Self-referential persistence loop: Continues iterating when the task is not yet complete, including ensuring a “completion promise” is produced.
Optional PRD mode: If you pass a PRD flag, it initializes a PRD file structure and breaks the work into right-sized, verifiable user stories before the loop starts.
Ultrawork parallel execution: Runs independent work in parallel using tiered model routing and background execution rules to reduce latency and save tokens.

Quick Start

Tell the AI: Continue solving my task with Ralph, and if needed use PRD mode when I include the flag --prd in my request.

Frequently Asked Questions about ralph

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

FAQPage Schema

How do I ensure iterative task execution reaches verified completion without partial stop conditions?▼

Iterative task execution reaches verified completion by running a self-referential persistence loop that continues iterating until an architect enforces review and demands fresh verification evidence, preventing premature or partial stop conditions.

How does parallel agent orchestration work with tiered model routing for code completion?▼

Parallel agent orchestration runs independent work simultaneously using ultrawork delegation, applying tiered model routing and background execution rules to coordinate tasks, reduce latency, and save tokens during code completion.

Can I initialize product requirements and break them into verifiable user stories before starting development?▼

Yes, you can initialize product requirements by passing a PRD flag. This creates a PRD file structure and automatically breaks the work into right-sized, verifiable user stories before the execution loop begins.

What is the best way to gate task completion in software development using architect verification?▼

The best way to gate task completion is enforcing an architect review that requires fresh verification evidence. This completion gating mechanism ensures work is proven complete before declaration, avoiding unverified partial deliveries.

Why does my AI agent stop working on a complex task before it is actually finished?▼

AI agents often stop early due to missing persistence loops. A structured persistence loop forces the agent to continue iterating when a task is incomplete, ensuring a completion promise is produced and verified.