Prompt Tool

Summary

The Prompt Tool enables the management, storage, and execution of prompts within the mindX system. Prompts are treated as first-class infrastructure, stored in memory, and can be versioned, shared, and executed by agents.

Technical Explanation

The Prompt Tool follows mindX doctrine:

• Memory is infrastructure: Prompts are persisted in memory and queryable
• Prompts are executable interfaces: Prompts can be executed with variable substitution
• Versioning and lineage: Prompts can be versioned and tracked through their evolution

Architecture

• Storage: Prompts are stored in data/prompts/ with a registry in prompt_registry.json
• Metadata: Each prompt has rich metadata including type, category, tags, version, and usage statistics
• Memory Integration: All prompt operations are logged to the Memory Agent for auditability

Prompt Types

• system: System-level prompts (inception, core instructions)
• agent: Agent-specific prompts
• user: User-defined prompts
• template: Reusable prompt templates with variables
• inception: Platform initialization prompts
• instruction: Instruction sets for agents

Categories

Prompts are organized by mindX layers:

• marketing: Marketing and narrative prompts
• community: Community and governance prompts
• development: Development and deployment prompts
• cognition: Cognitive and reasoning prompts
• execution: Execution and orchestration prompts
• governance: Governance and alignment prompts

Usage

Creating a Prompt

from tools.prompt_tool import PromptTool
from agents.memory_agent import MemoryAgent

memory_agent = MemoryAgent()
prompt_tool = PromptTool(memory_agent=memory_agent)
result = await prompt_tool.execute(
    operation="create",
    name="mindX Inception",
    content="You are mindX, an augmentic deployment platform...",
    description="Core mindX inception prompt",
    prompt_type="inception",
    category="governance",
    tags=["core", "inception", "platform"],
    author="system"
)

Retrieving a Prompt

result = await prompt_tool.execute(
    operation="get",
    prompt_id="prompt_id_here"
)

content = result["content"]
metadata = result["metadata"]

Executing a Prompt with Variables

result = await prompt_tool.execute(
    operation="execute",
    prompt_id="template_prompt_id",
    variables={
        "agent_name": "MyAgent",
        "task": "Analyze data"
    }
)

executed_content = result["executed_content"]

Searching Prompts

result = await prompt_tool.execute(
    operation="search",
    query="inception",
    search_content=True
)

prompts = result["results"]

Ingesting from External Sources

result = await prompt_tool.execute(
    operation="ingest",
    name="AgenticPlace Prompt",
    content="...",
    source="AgenticPlace",
    prompt_type="agent",
    category="development"
)

Versioning Prompts

result = await prompt_tool.execute(
    operation="version",
    prompt_id="original_prompt_id",
    new_content="Updated prompt content...",
    version_notes="Added new instructions"
)

Operations

• create: Create a new prompt
• get: Retrieve a prompt by ID
• update: Update an existing prompt
• delete: Delete a prompt
• list: List all prompts (with optional filters)
• execute: Execute a prompt with variable substitution
• search: Search prompts by content or metadata
• version: Create a new version of a prompt
• ingest: Ingest a prompt from external source

Integration

The Prompt Tool integrates with:

• Memory Agent: All operations are logged to memory
• BDI Agents: Agents can retrieve and execute prompts
• AgenticPlace: Prompts can be ingested from external sources

File Structure

data/prompts/
├── prompt_registry.json    # Metadata registry
└── {prompt_id}.prompt      # Individual prompt files

---