ralph

Spawn dedicated subagents per task across a six-phase workflow.

1|Updated Mar 6, 2026
One-click install
npx skills add https://github.com/adiman9/mnemos --skill ralph-adiman9
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ralph
Source: https://github.com/adiman9/mnemos/tree/main/core/skills/ralph
Command: npx skills add https://github.com/adiman9/mnemos --skill ralph-adiman9

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Processing AI agent tasks at scale requires strict isolation and reliable orchestration. Ralph provides a self-contained workflow that spawns a dedicated subagent per task and per phase to ensure fresh context and prevent cross-task contamination.

The system coordinates a multi-phase pipeline (extract, create, enrich, reflect, reweave, verify) with queue-driven progression, batch handling, and cross-sibling linkage to maintain traceability across related claims.

By enforcing end-to-end, deterministic behavior and standardized handoffs, it enables predictable outcomes and easier debugging in complex agent pipelines.

Core Features & Use Cases

  • Subagent-per-task phase isolation to prevent context leakage.
  • Serial and parallel processing modes with strict phase progression.
  • Batch-aware reflect/reweave and cross-link validation for consistency.
  • Mandatory spawn of subagents for every task and RALPH HANDOFF outputs for chainable pipelines.
  • Automatic queue advancement and per-claim note creation across phases.

Quick Start

Process the next N tasks from the queue in serial mode, spawning a subagent for each task.

Frequently Asked Questions about ralph

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

FAQPage Schema
How do I prevent context leakage when running multiple AI agent tasks in a pipeline?

Subagent orchestration prevents context leakage by spawning a dedicated subagent for every task and phase, ensuring fresh context isolation and preventing cross-task contamination during complex AI agent workflows.

What is the best way to orchestrate multi-phase AI agent workflows with strict task isolation?

Multi-phase AI agent orchestration coordinates a six-phase pipeline (extract, create, enrich, reflect, reweave, verify) with queue-driven progression and mandatory subagent spawning to enforce deterministic behavior and predictable outcomes.

Can I process task queues in parallel while maintaining traceability across related claims?

Queue processing supports serial and parallel modes with batch-aware reflect and reweave phases, utilizing cross-sibling linkage to maintain traceability and consistency across related claims during parallel task execution.

How do I chain AI agent pipelines together using standardized handoffs?

Pipeline chaining uses RALPH HANDOFF payloads to connect workflows, providing standardized handoff outputs that enable deterministic behavior and easier debugging across complex, multi-stage AI agent processing pipelines.

Does subagent task isolation work without external dependencies for complex agent orchestration?

Subagent task isolation operates as a self-contained workflow with no external dependencies, managing queue advancement, batch handling, and per-claim note creation entirely through its internal phase progression system.