For the complete documentation index, see llms.txt. This page is also available as Markdown.

Structured Hyperflow Prompt Editor

Build and refine agent behavior using a structured, section-based editor. No prompt engineering required.

The Structured Hyperflow Prompt Editor replaces the freeform prompt field with a guided authoring experience. Instead of writing and maintaining a single large prompt, you define agent behavior through dedicated sections, each focused on a specific aspect of how the agent should think and respond. The Prompt Editor compiles these inputs into a production-ready prompt configuration using standardized prompt engineering practices.

This approach is available for Hyperflows created after the feature is enabled. Hyperflows created before this feature continue to open in the standard editor and are not affected.

Before you begin: The Hyperflow Prompt Editor requires a feature flag to be enabled. See Enable Guided Agent Authoring

How it works

When you create a new Hyperflow, you describe the agent's intended behavior in plain language. The Prompt Editor uses that description to generate an initial configuration across five structured sections. You can then refine each section independently, optimize content with AI assistance, and preview the compiled prompt at any time before publishing.

The five sections that make up a Hyperflow prompt configuration are:

  • Agent Persona: Defines who the agent is and how it communicates with users.

  • Agent Responsibilities: Specifies the agent's objectives, capabilities, and boundaries.

  • Execution Plan: Describes the reasoning process and approach the agent uses to complete tasks.

  • Completion Criteria: Defines what constitutes a successful outcome or response.

  • Example Interactions: Provides representative scenarios and expected behaviors to guide agent performance.

Each section uses Markdown for content, supports independent editing, and can be expanded or collapsed in the accordion layout.

Two editors, two Hyperflow types

Hyperflow supports two prompt editors. The editor that opens for a given Hyperflow depends on which schema that Hyperflow uses.

Standard editor
Structured Prompt Editor

Authoring experience

Single freeform prompt field

Structured sections with dedicated purpose

Prompt writing required

Yes

No. Describe behavior in plain language.

AI assistance

Optimize entire prompt

Generate from description; Optimize entire prompt

Content format

HTML

Markdown

Governance

Manual

Built-in standardization across all agents

Hyperflows open in the editor that matches their schema. The two editors coexist and each Hyperflow always opens in its correct editor.

Last updated

Was this helpful?