multi-agent-guardrails-workflow

Explicit Multi-Agent Governance Designer

Moving from a single prompt to a complex multi-agent system introduces significant operational risk. This skill solves the "black box" orchestration problem by transforming your high-level system requirements into a deterministic, production-ready Mesh Flow guardrail topology.

Instead of relying on LLMs to "behave" via prompting, this tool generates hard-coded safety architectures. It classifies your agents' authority, defines strictly typed verifier contracts, and maps out explicit recovery paths for when agents fail. It bridges the gap between a whiteboard architectural concept and a deployable execution plan.

What it does

• Risk Classification: Automatically categorizes actions into observe-only, tool-write, or irreversible categories.
• Topology Generation: Emits a project.yaml identifying exactly where hard gates and human-in-the-loop approvals must occur.
• Contract Definition: Scripts technical verifier contracts that define pass/fail evidence requirements for every high-risk step.
• Artifact Production: Delivers structured JSON for integrations, Markdown for compliance reviews, and YAML for Mesh Flow runners.

Why use this skill

Prompting an AI to "follow safety rules" is unreliable. This skill provides a deterministic framework that forces your agents through a structured DAG (Directed Acyclic Graph) where safety is handled by the orchestration layer, not just the model's instructions. It is essential for teams working in high-stakes environments like FinTech, Healthcare, or CRM automation where data integrity is non-negotiable.