sddg

Coordinates multi-agent planning, implementation, and review cycles for complex software features.

Updated Nov 21, 2025
One-click install
npx skills add https://github.com/alatyshau/duet --skill sddg-alatyshau
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sddg
Source: https://github.com/alatyshau/duet/tree/main/packages/skills/workflows/sddg
Command: npx skills add https://github.com/alatyshau/duet --skill sddg-alatyshau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Complex features often fail when a single AI agent plans, codes, and reviews its own work without independent oversight. This Skill structures a four-persona workflow (Socrates, Daedalus-1, Daedalus-2, Hephaestus) that separates planning, implementation, and independent review into disciplined cycles with clear file-based handoffs. ## Core Features & Use Cases - Multi-Agent Review Cycle: Two independent reviewers examine code without reading each other's findings, then a coordinator synthesizes all issues into a single fix list. - File-Based Context Passing: Uses CurrentReview.md and topic files to transfer review feedback, test plans, and step statuses between agents working in separate sessions. - Spec-Driven Integrity: Enforces that spec/ remains the source of truth and that every commit contains code and spec updates together. - Use Case: When building a complex feature like a new authentication flow, use this workflow to have one agent plan the steps, another implement them, two reviewers independently audit the result, and a human run the final test plan before marking the step done. ## Quick Start Start the SDDG workflow for my current feature by creating a topic file with a plan, then coordinate implementation and independent review cycles until the test plan passes.

Frequently Asked Questions about sddg

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

FAQPage Schema
How do I coordinate multiple AI agents on a code review?

Use a file-based coordination pattern where reviewers write findings into separate sections of a shared CurrentReview.md file without reading each other's work. A coordinator agent then synthesizes all findings into a single fix list for the implementer.

What is the SDDG workflow for software development?

SDDG is a four-persona workflow: Socrates coordinates requirements and synthesis, Daedalus-1 plans architecture, Daedalus-2 reviews independently, and Hephaestus implements code and fixes. Steps cycle through review until approved, then a human runs the test plan.

Can AI agents mark development steps as complete?

No, in this workflow agents never mark steps as DONE. Only a human marks completion after running the test plan from CurrentReview.md, ensuring independent verification of the implementation.

How do reviewers stay independent in multi-agent review?

Each reviewer writes findings in their own section of CurrentReview.md and does not read other reviewers' sections before the coordinator synthesizes. This prevents anchoring bias and surfaces more distinct issues.

When should I not use a multi-agent review workflow?

Skip it for simple tasks handled by a solo agent with self-review, since the coordination overhead of topic files and review cycles adds no value. Reserve it for complex features requiring planning and independent verification.