conductor

Validate, run, and debug multi-agent workflows defined in YAML files.

Updated Mar 17, 2026
One-click install
npx skills add https://github.com/PolyphonyRequiem/twig --skill conductor-polyphonyrequiem
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: conductor
Source: https://github.com/PolyphonyRequiem/twig/tree/main/.github/skills/conductor
Command: npx skills add https://github.com/PolyphonyRequiem/twig --skill conductor-polyphonyrequiem

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires conductor, uv, and includes references (resource) components.

What problem does it solve? Orchestrating multiple AI agents through coordinated pipelines requires managing routing, context sharing, parallel execution, and human approvals manually. This Skill provides a CLI-driven way to validate, execute, and debug YAML-defined multi-agent workflows using the GitHub Copilot SDK or Anthropic Claude. ## Core Features & Use Cases - Workflow Execution: Run YAML-defined agent workflows with sequential, parallel, for-each, and loop-back routing, plus human-in-the-loop gates. - Validation & Debugging: Validate workflow schemas before execution, trace runs with verbose logging, and visualize execution in a real-time web dashboard. - Context & Cost Management: Control how agents share data (accumulate, last_only, explicit modes) and track per-agent token usage and costs. - Use Case: Run the pre-built document workflow to produce a research report that passes through an author agent and two independent review cycles (content and quality) until both score above 85. ## Quick Start Run the document workflow by asking the AI to execute the conductor workflow with a purpose input describing the document you want created.

Frequently Asked Questions about conductor

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

FAQPage Schema
How do I run a multi-agent workflow with conductor?

Run a workflow with 'conductor run workflow.yaml --input key="value"' in a background terminal, since executions typically take around 15 minutes. Use the -V flag for verbose output showing full prompts and tool calls, or --web for a real-time dashboard.

How do I validate a conductor workflow YAML file before running it?

Run 'conductor validate workflow.yaml' to check YAML syntax, required fields, agent references, route reachability, and template syntax without executing. You can also use --dry-run to preview the execution plan.

Does conductor support both Copilot and Claude providers?

Yes, conductor supports the GitHub Copilot SDK as the default provider and Anthropic Claude as an alternative. Set the provider at the workflow level or override it per agent, and use -p claude on the CLI to switch providers for an entire run.

Why does conductor fail on Windows with 'No module named fcntl'?

This is a known regression in conductor's copilot provider where _fix_pipe_blocking_mode() unconditionally imports the Unix-only fcntl module. Run 'conductor update' to get the latest version; if the error persists, the fix has not been released upstream.

How do I add human approval steps to an AI agent workflow?

Define an agent with type human_gate, a prompt, and a list of options where each option has a label, value, and route. The workflow pauses for user selection, and you can collect text feedback with prompt_for or bypass gates using --skip-gates.