Skip to main content

​
Overview

The Prompts domain manages the SO1 platform’s AI prompt engineering lifecycle through the Veritas prompt library. These agents optimize prompt design, build multi-step reasoning chains, and curate reusable prompt fragments for consistent AI behavior across all platform operations.
Veritas Integration: All agents in this domain interact directly with the Veritas prompt library (prompts/veritas/) to ensure prompts follow schema standards, maintain version control, and integrate seamlessly with the Factory Orchestrator.

​
Agents in This Domain

Prompt Refiner

Optimizes individual prompts for clarity, effectiveness, and token efficiency

Chain Architect

Designs multi-step prompt chains for complex reasoning workflows

Fragment Curator

Manages reusable prompt fragments and template libraries

​
Domain Responsibilities

​
Prompt Engineering

  • Clarity optimization: Refine instructions for unambiguous AI interpretation
  • Token efficiency: Reduce prompt length while preserving effectiveness
  • Schema compliance: Ensure all prompts follow Veritas standards
  • Version control: Track prompt evolution and performance metrics

​
Chain Design

  • Multi-step reasoning: Break complex tasks into sequential prompt chains
  • State management: Handle context passing between chain steps
  • Error recovery: Design fallback strategies for chain failures
  • Performance optimization: Balance chain depth with execution speed

​
Library Management

  • Fragment reuse: Create modular prompt components for consistency
  • Template versioning: Manage prompt template lifecycle
  • Cross-agent sharing: Enable prompt reuse across different agents
  • Documentation: Maintain clear usage guidelines for each fragment

​
Veritas Library Structure

prompts/veritas/
├── agents/                 # Agent-specific prompts
│   ├── orchestration/
│   ├── automation/
│   └── ...
├── chains/                 # Multi-step prompt chains
│   ├── incident-response/
│   └── deployment/
├── fragments/              # Reusable prompt components
│   ├── common/            # Universal fragments
│   ├── domain/            # Domain-specific fragments
│   └── gates/             # FORGE gate validation fragments
└── templates/              # Base prompt templates
    ├── system/
    └── user/

​
Integration with FORGE

All prompt agents operate within the FORGE execution model to ensure gate compliance:
  • Review existing prompts for optimization opportunities
  • Analyze prompt performance metrics and failure patterns
  • Assess Veritas library structure and fragment usage
  • Create new prompts following Veritas schema
  • Design prompt chains with state management
  • Build reusable fragments for common patterns
  • Test prompts with representative inputs
  • Validate chain execution and state transitions
  • Verify fragment integration in target agents

​
Common Use Cases

​
Prompt Optimization

When an existing prompt produces inconsistent results or uses excessive tokens, the Prompt Refiner analyzes the prompt structure and suggests improvements for clarity and efficiency.

​
Complex Workflows

For tasks requiring multi-step reasoning (e.g., incident triage → severity assessment → mitigation planning), the Chain Architect designs sequential prompt chains with proper state management.

​
Cross-Agent Consistency

When multiple agents need similar instructions (e.g., “follow FORGE gates” or “output JSON schema”), the Fragment Curator creates reusable fragments that ensure consistent behavior.

​
Source Repository

View Source Code

All Prompts domain agents are maintained in the so1-agents repository under agents/prompts/