ralph

Execute persistent task loops with parallel delegation and architect verification.

Updated Apr 20, 2026
One-click install
npx skills add https://github.com/hotaq/Sprite_harmess --skill ralph-hotaq
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ralph
Source: https://github.com/hotaq/Sprite_harmess/tree/main/.codex/skills/ralph
Command: npx skills add https://github.com/hotaq/Sprite_harmess --skill ralph-hotaq

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Ralph solves the common failure mode where complex work stops too early by keeping a task moving through retries, verification, and cleanup until it is genuinely complete.

Core Features & Use Cases

  • Persistent execution loop: Continues multi-step work across iterations instead of treating partial progress as finished.
  • Parallel delegation and verification: Routes independent work in parallel and requires fresh evidence before completion.
  • Session-aware recovery: Tracks state, resumes from prior progress, and validates that work survives failures and retries.
  • Use case: Use Ralph when an engineering task must be finished end-to-end with architect approval, test evidence, and safe shutdown.

Quick Start

Ask Ralph to keep working on the task until it is fully verified and cleanly completed.

Frequently Asked Questions about ralph

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

FAQPage Schema
How do I keep an engineering workflow running until the task is fully verified?

To keep an engineering workflow running until verified, use a persistent task execution loop that continues multi-step work across iterations, applies repeated validation, and requires fresh evidence before completion.

What is session-aware recovery for complex automation workflows?

Session-aware recovery in automation workflows tracks session state, resumes tasks from prior progress, and validates that work survives failures and retries without losing partial results.

How do I manage parallel delegation and verification in a multi-step task?

Parallel delegation and verification routes independent work streams concurrently and requires fresh evidence gathering and tiered review before the overall task is considered complete.

When do I need a persistent execution loop for engineering tasks?

You need a persistent execution loop for engineering tasks when work must be finished end-to-end with architect approval, test evidence, and safe shutdown, preventing partial progress from being treated as finished.

Can I resume a failed task from its previous session state?

Yes, you can resume a failed task from previous session state by tracking recovery progress across iterations and validating that work survives failures before attempting retries.

What's the best way to ensure a complex task finishes with clean cleanup?

The best way to ensure a complex task finishes with clean cleanup is to enforce explicit evidence gathering, tiered architect review, and safe shutdown procedures before the execution loop terminates.