agent-builder

Design single-agent or multi-agent architectures for decomposed processes.

3|Updated Apr 5, 2026

Search Tags:#architecture #multi-agent #agents #spec #orchestrator #workflow-design #process-decomposition

Ready to run

npx skills add https://github.com/dvy1987/agent-loom --skill agent-builder-dvy1987

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agent-builder
Source: https://github.com/dvy1987/agent-loom/tree/main/.agents/skills/agent-builder
Command: npx skills add https://github.com/dvy1987/agent-loom --skill agent-builder-dvy1987

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill helps teams design robust execution architectures for decomposed processes, deciding between single-agent and multi-agent topologies and persisting architecture specs for learning loops.

Core Features & Use Cases

Read and interpret a decomposed process from docs/processes to determine structure.
Decide between single-agent orchestration and multi-agent topology with clearly defined agent boundaries and handoff protocols.
Persist architecture specifications to docs/architecture/ and update process entries with architecture references.
Provide prompts for agent design and orchestrator behavior, and handle complex handoffs and failures.

Quick Start

Provide an architecture design for the given process using the multi-agent guidelines.

Frequently Asked Questions about agent-builder

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

FAQPage Schema

How do I design a multi-agent architecture for a complex workflow?▼

To design a multi-agent architecture, you map decomposed processes with defined steps and boundaries into a topology. This generates role prompts, defines handoff protocols, and handles sequential, parallel, or hierarchical task graphs.

When should I choose a single agent versus a multi-agent topology?▼

Choose a single agent for simpler workflows or a multi-agent topology when processes require distinct boundaries and complex handoffs. The architecture design evaluates the decomposed process structure to determine the optimal orchestrator setup.

What is the best way to define handoff protocols between agents?▼

Defining handoff protocols requires establishing clear agent boundaries within the process decomposition. The architecture design generates specific prompts for agent design and orchestrator behavior to manage these complex handoffs and failures.

How do I persist agent architecture specifications for future learning loops?▼

To persist agent architecture specifications, save them to the docs/architecture directory. This ensures the specs are retained for learning loops and updates the original process entries with the corresponding architecture references.

Can I apply process decomposition to hierarchical task graphs?▼

Yes, process decomposition applies to hierarchical task graphs, as well as sequential and parallel structures. The architecture design interprets these defined steps from docs/processes to determine the appropriate agent boundaries and toolchains.