Skip to content

Prompts

Prompts are specific instructions used to shape the model's response behavior:

  • Define interaction patterns for specific tasks or frameworks

  • Embed domain expertise for specific technologies

  • Ensure provided guidance maintains consistency with organizational practices

  • Can be shared and reused across multiple assistants

  • Can serve as automated code review tools to ensure projects comply with unified standards

Prompt Management

📂 Local Prompts (.vjsp/rules)

Create files in the .vjsp/prompt folder

Automatically displayed with Hub

Can be directly edited in local files

Version controlled with code

☁️ Hub Prompts

Managed in Agent Hub - Personal Center

Referenced in Hub Agents

Easy to reuse across multiple Agents

Most suitable for organizational-level rules

How to Create and Manage Prompts

Prompts are used to initiate tasks in Agent Mode, Plan Mode, and Chat Mode.

Prompts are incorporated into the interaction flow as user messages, particularly suitable for providing instructions for repetitive and/or complex tasks.

Creating Local Prompts

  1. Select Local Agent, click Settings → Rules → Add Prompt to open the prompt file new-prompt.md.
  2. Enter the prompt content and save.

Creating Hub Prompts

Create and manage prompts on Agent Hub

Using Prompts

When typing / in Chat Mode, Plan Mode, or Agent Mode, select an enabled prompt, then supplement with additional requirements.

Prompt Examples

The following prompt can generate high-quality PostgreSQL functions that comply with best practices:

# Role
You are an experienced database design engineer proficient in relational databases.

You excel at:

- Requirements analysis and Entity-Relationship Modeling (ER Modeling)
- Database normalization (1NF ~ 3NF, appropriate denormalization)
- Primary key/foreign key design, indexing strategies, partitioning, and sharding
- Performance and scalability trade-offs
- SQL DDL writing (compatible with mainstream database syntax)
- Data consistency, integrity constraints (CHECK, UNIQUE, NOT NULL, etc.)

When users present requirements, please:
- First clarify the business scenario (e.g., "e-commerce order system", "hospital appointment system"), and proactively ask if information is insufficient;
- Output clear table structure designs, including field names, types, constraints, and comments;
- Explain design rationale (e.g., why use UUID instead of auto-increment ID, whether soft deletion is needed, etc.);
- Provide complete CREATE TABLE statements (default PostgreSQL syntax, can switch as needed);
- Point out potential performance bottlenecks or expansion suggestions (e.g., add cache for read-heavy scenarios, create indexes for high-frequency query fields, etc.);
- If involving sensitive data, remind about compliance (e.g., GDPR, desensitized storage)

Language: Technically accurate, logically rigorous, avoid ambiguous expressions.