rat-dse

Coordinates Design, Space Exploration across phases with git worktrees and self-critique.

43|9|Updated Feb 25, 2026
One-click install
npx skills add https://github.com/babyworm/rtl-agent-team --skill rat-dse
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: rat-dse
Source: https://github.com/babyworm/rtl-agent-team/tree/main/skills/rat-dse
Command: npx skills add https://github.com/babyworm/rtl-agent-team --skill rat-dse

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This DSE skill enables deep, iterative exploration of multiple architecture and algorithm candidates, including a self-critique loop and user-guided trial iterations, to converge on a strong pre-implementation plan.

Core Features & Use Cases

  • Self-critique loop: automatically reviews its own output and replans Phase 1→3 with findings.
  • Multi-phase orchestration: handles Phase 1 (research + algorithm study), Phase 2 (architecture exploration + reference C model), and Phase 3 (μArch + BFM) to produce a complete design package.
  • Trial isolation via worktrees: uses git worktrees to compare alternate designs safely without compromising the current baseline.
  • Deliverables: artifacts like phase documents, decisions logs, and a ready pre-implementation package for downstream RTL integration.

Quick Start

Start DSE Trial 1 by providing your spec or project description; the orchestrator will run Phase 1–3, perform self-critique, and guide iterative trials until you approve.

Frequently Asked Questions about rat-dse

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

FAQPage Schema
How do I perform iterative Design Space Exploration for RTL design with multiple architecture candidates?

Iterative Design Space Exploration runs phases from algorithm study to microarchitecture design, using a self-critique loop to review outputs and replan until a strong candidate converges, yielding a pre-implementation package for downstream RTL integration.

What is the best way to compare alternate architecture candidates without compromising my current baseline?

Comparing alternate architecture candidates safely uses git worktrees for trial isolation, managing isolated worktrees to explore multiple design variations in parallel without compromising the current project baseline.

How does self-critique work in architecture and microarchitecture exploration workflows?

Self-critique in architecture exploration automatically reviews generated phase outputs and replans design phases with findings, continuously evaluating artifacts and guiding iterative trials until the user approves the final design.

Do I need a reference C model to start design space exploration for my hardware project?

A reference C model is supported during Phase 2 architecture exploration to help evaluate candidates; you can start by providing your spec or project description, and the orchestrator handles algorithm studies and reference model generation.

What deliverables should I expect from an automated DSE workflow before starting RTL implementation?

Deliverables from a DSE workflow include phase documents, decision logs, and a ready pre-implementation package, capturing explored architecture and microarchitecture decisions to directly feed downstream RTL integration.