openserv-multi-agent-workflows_skill

Overview

This skill provides hands-on multi-agent workflow examples for the OpenServ Platform, demonstrating how to compose, orchestrate, and monitor collaborative agent pipelines. It focuses on agent discovery, workspace creation, task dependencies, and explicit workflow edges to ensure reliable execution. Use the included examples and reference material to build production-grade multi-agent automation.

How this skill works

The examples show how to authenticate with the platform client, discover marketplace agents, and provision workflows with triggers, tasks, and edges. Each task is bound to an agent and can declare dependencies; workflow edges explicitly connect triggers and tasks so outputs flow correctly. The package includes patterns for sequential pipelines, fan-out/combiner stages, and conditional branching using outputOptions and sourcePort.

When to use it

• You need coordinated work across multiple specialized agents (research → write → image).
• Building pipelines with clear task dependencies and deterministic data flow.
• Prototyping multi-agent orchestration patterns before production deployment.
• When you want examples for agent discovery and marketplace integration.
• Creating conditional or parallel branches with a final combiner task.

Best practices

• Always define workflow edges; task dependencies alone do not connect nodes.
• Choose a polished workflow name and a detailed goal sentence for provisioning.
• Start linear, add complexity only when necessary to avoid tangled graphs.
• Give each task a single, clear purpose and minimize cross-connections.
• Use one combiner task to merge parallel branches and keep outputs coherent.

Example use cases

• Blog pipeline: Research agent gathers sources, writer agent drafts the post.
• Content creation: Research → write → image generation pipeline for social posts.
• YouTube-to-blog: Transcribe, summarize, and expand video content into articles.
• Life coaching workflow: Multi-step intake, analysis, personalized plan, and follow-up.
• Analytics pipeline: Data extraction → analysis agent → report generator.