Gendial

English

简体中文

English

简体中文

Appearance

Digital Employee Configuration ​

A Digital Employee's behavior is defined by three core documents that collectively form its "brain". During each conversation, these three documents are combined into the system prompt, directly determining the Digital Employee's reply style and behavioral patterns.

Overview ​

The configuration entry is located under the Config menu item in the Digital Employee detail page. The page contains three tabs, each corresponding to a Markdown document.

The design of these three documents draws from the Persona Engineering philosophy in the AI Agent field: decomposing an Agent's identity into three independent dimensions—"what to do", "who I am", and "how to express"—allowing each behavioral adjustment to be made independently without interfering with others.

AGENTS.md — Behavioral Guidelines ​

AGENTS.md answers the question: What does this Digital Employee do and how does it work.

It serves as the Digital Employee's "job description". Typical content includes:

  • Scope of Responsibility: What types of requests the Digital Employee can handle and cannot handle
  • Operational Standards: Procedures and constraints to follow when executing tasks
  • Reply Style: Tone requirements, format specifications, content boundaries
  • Collaboration Conventions: Interaction style recommendations when working with users

For example, a finance Digital Employee's AGENTS.md might specify: "Only answer finance-related questions. When encountering tax inquiries, suggest contacting a professional tax advisor."

SOUL.md — Self-Awareness ​

SOUL.md answers the question: What kind of person is this Digital Employee.

It endows the Digital Employee with a consistent "personality", maintaining a stable sense of identity across multiple conversations with users. Typical content includes:

  • Mission Statement: The purpose and value of the Digital Employee's existence
  • Service Style: Style positioning such as professional, approachable, concise
  • Behavioral Commitments: Commitments to users regarding confidentiality, response speed, accuracy

The design philosophy of SOUL.md: even with the same workflow and knowledge base, different SOULs produce completely different interaction experiences—one could be a professional and rigorous "technical advisor", another a friendly and lively "learning assistant".

IDENTITY.md — Identity Statement ​

IDENTITY.md answers the question: Who is this Digital Employee externally.

It clarifies the Digital Employee's identity boundaries. Typical content includes:

  • Identity Positioning: The Digital Employee's role name and position
  • Capability Boundaries: Clearly informing users what the Digital Employee can and cannot do
  • Privacy and Security Declarations: Data usage policies and limitation statements

The role of IDENTITY.md is to set user expectations at the start of the conversation, preventing users from making requests beyond the Digital Employee's capabilities.

How the Three Documents Work Together ​

At the start of each conversation, the system combines the three documents in order as the system prompt:

[AGENTS.md content]
---
[SOUL.md content]
---
[IDENTITY.md content]

This structure follows a logical progression from macro to micro, from functionality to personality. After receiving the system prompt, the LLM generates appropriate responses based on these three layers of constraints.

After modifying any document, click Save for changes to take effect in the next conversation. It is recommended to test via the chat function after modifications to verify that behavioral changes meet expectations.

Last updated: