Decompose complex user instructions into a structured multi-agent execution graph.
npx skills add https://github.com/dewitt/swarm --skill planning-agent-dewitt
Please help me install this Agent Skill. Skill: planning_agent Source: https://github.com/dewitt/swarm/tree/main/skills/planning-agent Command: npx skills add https://github.com/dewitt/swarm --skill planning-agent-dewitt
Planning complex user requests manually is error-prone and time-consuming; this Skill provides a principled way to decompose tasks into an executable multi-agent plan with persistent state coordination.
Provide a complex user task and planning_agent will generate a multi-agent execution graph to accomplish it.
High-intent search queries and answers about installing and using this skill.
Multi-agent planning decomposes complex instructions into a structured execution graph across specialized agents. You need it when tasks require cross-agent collaboration, persistent session state, and deterministic planning with explicit dependencies to ensure accurate execution.
You decompose requests by providing a complex task to a planning agent, which generates a strict DAG with named spans and a terminal synthesis node. It outputs a defined JSON schema coordinating sequential and parallel agent spans across specialized roles.
Yes, you can maintain persistent session state to preserve context across turns. The decomposition process coordinates specialized Swarm agents by enforcing a strict DAG with named spans, ensuring context is retained throughout the execution graph.
Yes, multi-agent planning works with both parallel and sequential execution spans. It decomposes complex goals into specialized agent roles within a strict DAG, coordinating cross-agent collaboration while maintaining persistent session state throughout the workflow.
A strict DAG imposes deterministic planning with explicit dependencies and a terminal synthesis node, meaning it does not support dynamic cyclic routing or runtime loop modifications. Complex goals must fit within defined safety constraints and a static execution graph structure.
ntholm86Analyze user prompts to extract core intentions using session context.
ZealousEarOrchestrate multiple Codex agents to parallelize coding tasks and aggregate results.
CornjebusOrchestrate distributed research, development, and testing workflows with multi-agent swarm topologies.
quanngynxOrchestrate durable multi-step workflows with Vercel Workflow DevKit.
