Coordinate multi-agent design debate rounds and write final synthesis.
npx skills add https://github.com/slior/dialectic-agentic --skill orchestrate-slior
Please help me install this Agent Skill. Skill: orchestrate Source: https://github.com/slior/dialectic-agentic/tree/main/.cursor/skills/orchestrate Command: npx skills add https://github.com/slior/dialectic-agentic --skill orchestrate-slior
Coordinates and automates end-to-end multi-agent design debates, coordinating configuration, workspace setup, and round-based evaluation to produce a converged solution.
Run the orchestrator with the required WORKSPACE and PROJECT paths to start a new multi-agent debate.
High-intent search queries and answers about installing and using this skill.
A multi-agent design debate workflow coordinates agents through proposal, critique, and refinement rounds to converge on a synthesized design solution. It automates configuration loading, workspace setup, and judge-based convergence checks to produce structured outputs.
To orchestrate multi-agent debates end-to-end, run the orchestrator with required WORKSPACE and PROJECT paths. It initializes directories, runs optional clarifications, executes round phases, dispatches a judge for convergence, and writes the final synthesis automatically.
Yes, prompt-dispatch supports assigning multiple agent roles within the orchestration workflow. Agents execute structured proposal, critique, and refinement phases, while a central judge evaluates convergence and enforces guardrails throughout the debate lifecycle.
No external dependencies are required to run multi-agent orchestration. The workflow operates standalone using internal scripts to load configurations, initialize workspaces, manage debate directories, validate outputs, and execute recovery steps without additional packages.
The best way to check convergence is dispatching a central judge agent that evaluates proposal alignment after refinement rounds. The orchestrator enforces this convergence check, writing status.json and progress.md files to track validation and recovery steps.
Structured outputs like status.json and progress.md are needed to enforce guardrails, track round phases, and enable validation and recovery steps during multi-agent debates. They provide persistent state tracking across proposal, critique, and refinement cycles for final synthesis.
oaustegardInitialize Git repositories and commit changes for skill directories.
kelsi-andrewssDebate topics adversarially between Claude and Gemini and generate a final synthesis.
sliorExecute one debate phase for a role-specific subagent in a multi-agent design debate.
northseadlDesign, simulate, and evaluate LLM agents with pattern-matrix mode selection.
