dyad:swarm-to-plan

Orchestrates PM, UX, and engineering agents to debate an idea and write a spec to plans/.

21.4k|2.6k|Updated Apr 11, 2025
One-click install
npx skills add https://github.com/dyad-sh/dyad --skill dyad-swarm-to-plan
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: dyad:swarm-to-plan
Source: https://github.com/dyad-sh/dyad/tree/main/.claude/skills/swarm-to-plan
Command: npx skills add https://github.com/dyad-sh/dyad --skill dyad-swarm-to-plan

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Turning a rough feature idea into an actionable engineering spec requires balancing product, design, and technical perspectives, which is hard to do in a single pass. This Skill runs a structured multi-agent planning session that surfaces ambiguities, resolves trade-offs with the human, and produces a complete plan document.

Core Features & Use Cases

  • Multi-Agent Debate: Spawns Product Manager, UX Designer, and Engineering Lead agents that analyze the idea independently, then challenge and build on each other's analyses.
  • Human Clarification Loop: Synthesizes open questions from the debate and asks the user up to four prioritized clarifying questions before finalizing.
  • Structured Spec Output: Writes a comprehensive plan to plans/<plan-name>.md covering scope, user stories, UX flows, technical design, implementation phases, testing, and risks.
  • Use Case: Given an idea like "add collaborative editing", the team debates scope and architecture, asks you about MVP boundaries, then delivers a phased implementation spec with a decision log.

Quick Start

Run the swarm-to-plan skill with the argument "redesign the settings page" to start a planning session that ends with a spec written to the plans directory.

Frequently Asked Questions about dyad:swarm-to-plan

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

FAQPage Schema
How do I turn a feature idea into an engineering spec with AI agents?

Pass the idea as an argument to this skill, which spawns PM, UX, and engineering agents to analyze and debate it. After a clarification round with you, it writes a full spec with scope, user stories, technical design, and implementation phases to plans/<plan-name>.md.

What roles participate in the swarm planning session?

Three agents participate: a Product Manager covering problem definition and scope, a UX Designer covering user flows and accessibility, and an Engineering Lead covering architecture, data models, and risks. Their role definitions live in the references directory.

Where is the final plan written and what does it contain?

The plan is written to plans/<plan-name>.md in the repository. It includes a summary, problem statement, MVP scope, user stories, UX design, technical design, phased implementation checklist, testing strategy, risk table, open questions, and a decision log.

What happens if a planning agent fails to respond?

The session proceeds with the remaining agents' input and still writes the plan file. If the user declines to answer clarifying questions, the team proceeds with its best assumptions and notes them in the plan.

When should I use swarm planning instead of writing a spec directly?

Use it for ideas with genuine cross-functional ambiguity where product scope, UX trade-offs, and technical feasibility interact. For small, well-understood changes, the multi-agent debate and clarification rounds add overhead without proportional benefit.