pi-flow

Execute deterministic multi-agent workflows defined in JSON and Markdown.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/choru-k/skills-for-ai --skill pi-flow
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: pi-flow
Source: https://github.com/choru-k/skills-for-ai/tree/main/public/pi/pi-flow
Command: npx skills add https://github.com/choru-k/skills-for-ai --skill pi-flow

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill provides a robust engine for defining, executing, and inspecting complex multi-agent workflows, moving beyond simple agent-to-agent communication to structured, deterministic, and resumable process orchestration.

Core Features & Use Cases

  • Workflow Definition: Define complex process topologies using strict JSON (flow.json) and agent behavior in Markdown (agents/*.md).
  • Deterministic Execution: Run workflows step-by-step (flow.step) or until completion (flow.run.until_wait), ensuring predictable outcomes.
  • Resumable Runs: Persist runtime state, allowing workflows to be paused and resumed, ideal for long-running tasks or interruptions.
  • Use Case: Automate a software development lifecycle, from planning and code implementation to review and deployment, coordinating multiple specialized agents (planner, reviewers, workers) through a defined workflow.

Quick Start

Use the pi_flow skill to start a new workflow run named 'code-review-process' using the 'code-review-workflow' definition.

Frequently Asked Questions about pi-flow

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

FAQPage Schema
How do I orchestrate multi-agent workflows with deterministic execution?▼

You can orchestrate multi-agent workflows by defining process topology in strict JSON and agent behavior in Markdown, executing them step-by-step or until completion for deterministic outcomes.

Can I pause and resume long-running agent coordination tasks?▼

Yes, agent coordination tasks support resumable runs through runtime state persistence, allowing workflows to be paused and resumed during long-running tasks or interruptions.

How do I define complex agent coordination patterns for a software development lifecycle?▼

Complex agent coordination patterns are defined using strict JSON for workflow topology and Markdown files for agent behavior, enabling specialized agents like planners and reviewers to collaborate.

Does pi-flow support inspecting multi-agent workflow state during execution?▼

Yes, workflow state can be inspected during execution using a live TUI inspector, allowing users to monitor agent spawning, message routing, and artifact validation in real-time.

What is the best way to validate results and artifacts in a multi-agent workflow?▼

The best way to validate artifacts is by using result contracts within the workflow definition, ensuring deterministic execution and predictable outcomes across coordinated agents.

Can I run multi-agent workflows step-by-step instead of all at once?▼

Yes, workflows can be executed step-by-step using specific execution commands, ensuring deterministic execution and allowing users to inspect agent state before proceeding to the next phase.