All products

Map an agent workflow with AI

Describe an agent system in editable YAML, then add a scenario showing how one request moves between its parts.

Map an agent workflow with AI when a box labelled “assistant” hides too many moving parts. A support request might involve a user, an orchestrating agent, a search tool, a model and a knowledge base. The agentic-diagrams skill gives each part a stable identifier and describes the connections in YAML.

The diagram can also contain a scenario. Follow one request through the existing nodes and show where it branches, returns a result or stops. A bundled schema lets the agent check the file locally. You decide separately whether to import the result into the maker's hosted viewer.

What it does

  • Map the tools and models an agent actually calls.
  • Distinguish an orchestrator from the agents it delegates to.
  • Show where a knowledge base or memory store is consulted.
  • Walk through a successful request and a blocked-input scenario.
  • Check identifiers and permitted field names against the included schema.

Install and use

  1. Give the agent a component list and one representative request. Authoring requires neither an account nor a package installation.
  2. Review the nodes and connections, then add the sequence for a real scenario.
  3. Check the YAML against the bundled reference. Import it into the hosted viewer only if you intend to share that information there.

FAQ

Source

Made by Paul Kohler (Agentic Diagrams). Original skill: agentic-diagrams, published on GitHub under the MIT licence.

View on GitHub