swarm-to-plan_skill

This skill coordinates a PM, UX, and engineering swarm to debate ideas and generate a detailed plan written to plans/$plan-name.md.
  • TypeScript

19.7k

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill dyad-sh/dyad --skill swarm-to-plan

  • SKILL.md8.6 KB

Overview

This skill runs a structured swarm planning session using three specialist agents (Product Manager, UX Designer, Engineering Lead) to debate an idea, surface ambiguities, ask clarifying questions, and produce a detailed implementation plan saved to plans/<plan-name>.md. It automates team creation, role-driven analysis, cross-role debate, prioritized questions for the human, and compilation of a final spec that aligns with product principles from rules/product-principles.md.

How this skill works

The skill reads the idea from the input argument and scans the repo for tech stack and patterns (package.json, configs, key files) to build a short codebase context summary. It spawns three subagents with role-specific prompts, collects initial analyses, facilitates a structured debate, compiles prioritized clarifying questions for the user, ingests answers, gathers final refinements, synthesizes a single coherent plan, writes plans/<plan-name>.md, and shuts down the temporary team.

When to use it

  • Turning an early idea or feature request into a concrete implementation plan
  • Preparing a scoped MVP spec before engineering work begins
  • Aligning product, UX, and engineering on trade-offs and unanswered questions
  • Needing a reproducible, auditable planning output written to the repo
  • When you want the team to follow established product design principles automatically

Best practices

  • Provide a concise, specific idea description as the argument to reduce low-value clarifications
  • Review rules/product-principles.md in the repo before running to ensure alignment
  • Answer the prioritized user questions promptly to minimize follow-up rounds
  • Use slug-friendly plan names (lowercase, hyphens) so the output file path is neat
  • Keep human responses focused on scope, priority, and constraints for highest impact

Example use cases

  • Plan an MVP for collaborative document editing across web and mobile
  • Design and scope a settings page redesign with accessibility requirements
  • Evaluate technical approach and rollout plan for adding real-time notifications
  • Create a phased implementation plan for migration to a new auth system
  • Generate a spec for adding analytics events and dashboard pages

FAQ

The skill proceeds with available analyses, notes the missing perspective in the plan, and still writes plans/<plan-name>.md. It flags the gap as an open risk.

How many questions will the team ask the human at once?

Up to four highest-impact questions per round. If more remain, additional rounds will be used after answers arrive.

Where is the final plan saved?

In the repository under plans/<plan-name>.md. The skill creates the plans directory if it doesn't exist.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
swarm-to-plan skill by dyad-sh/dyad | VeilStrat