pipeline

Orchestrate multi-stage agent workflows with resumable state persisted to JSON.

2|Updated May 11, 2026
One-click install
npx skills add https://github.com/xz1220/oh-my-kimi --skill pipeline-xz1220
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pipeline
Source: https://github.com/xz1220/oh-my-kimi/tree/main/skills/pipeline
Command: npx skills add https://github.com/xz1220/oh-my-kimi --skill pipeline-xz1220

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill removes the friction of manually coordinating multiple agent stages (planning, execution, and verification) by providing a single pipeline runner with resumable state.

Core Features & Use Cases

  • Configurable multi-stage orchestration: Run a sequence of stages through a unified PipelineStage interface, passing accumulated context between phases.
  • Built-in autopilot stages: Use an opinionated pipeline chaining RALPLAN consensus planning, team-based execution, and Ralph verification.
  • Resume via persisted state: Persist pipeline progress to .omk/state/pipeline-state.json and automatically continue from the last incomplete stage.
  • Use case: You need to turn a complex requirement into an implementable plan, execute it with multiple workers, and then validate results through an iterative verification loop.

Quick Start

Ask your agent to run a pipeline for build feature X and execute RALPLAN, then team execution with workerCount set to 3, and finally Ralph verification with maxIterations set to 15.

Frequently Asked Questions about pipeline

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

FAQPage Schema
How do I orchestrate multi-stage agent workflows with resumable state?

Multi-stage agent workflows with resumable state are orchestrated through a unified pipeline runner that checkpoints progress to .omk/state/pipeline-state.json. This allows automatically resuming execution from the last incomplete stage across runs.

What is consensus planning and how does it fit into agent orchestration?

Consensus planning is the initial stage of agent orchestration implemented via RALPLAN. It generates an implementable plan from complex requirements, passing accumulated context to downstream team execution and iterative verification phases.

How do I execute a plan with multiple workers in a pipeline?

Executing a plan with multiple workers involves running a team-based execution stage within the pipeline. You can configure the workerCount parameter, such as setting it to 3, to parallelize tasks before final verification.

Can I skip specific stages in a configurable multi-stage pipeline?

Yes, specific stages can be skipped in a configurable multi-stage pipeline. The system implements a PipelineStage interface that supports stage skipping, allowing you to bypass unnecessary phases while maintaining accumulated context.

Does iterative verification work with parallel team execution?

Iterative verification works directly with parallel team execution via Ralph verification. After team execution completes, the pipeline validates results through an iterative loop with configurable parameters like maxIterations set to 15.