Manual Setup

Works with any AI model: ChatGPT, Gemini, Claude (web), local LLMs, or any other.

Setup

1. Download the repository

   git clone https://github.com/kd5ziy/AIMantras.git

   Or download as ZIP from GitHub and extract.

2. Give your AI access to the files

   Depending on your AI interface:

   • File upload: Upload the entire Prompt-AI-Mantras/ folder and AIMantra.md
   • Copy/paste: Copy file contents into the conversation
   • Local LLM: Point to the directory in your context

Usage Methods

Full Context Load

For AI models with large context windows, load everything upfront:

I'm going to give you the AI Mantras framework. Please read and internalize these files:

1. First, read AIMantra.md - this is your entry point and triage guide
2. Read Prompt-AI-Mantras/principles/guiding-principles.md - internalize these values
3. You now have access to personas in Prompt-AI-Mantras/personas/
4. You have access to patterns in Prompt-AI-Mantras/patterns/
5. You have access to skills in Prompt-AI-Mantras/skills/

Follow the AIMantra.md instructions to triage my requests and load only what you need.

My request: [YOUR REQUEST HERE]

Guided Loading

For AI models with limited context, load incrementally:

Step 1: Load the entry point

Read this file and follow its instructions:
[Paste contents of AIMantra.md]

My request: [YOUR REQUEST HERE]

Step 2: The AI will tell you what else to load based on your request complexity. Provide files as requested.

Direct Persona

If you know which persona you need:

You are Clara, a Financial Analyst. Read and embody this persona:
[Paste contents of Prompt-AI-Mantras/personas/domain/Clara-Financial-Analyst.md]

Also read and internalize these guiding principles:
[Paste contents of Prompt-AI-Mantras/principles/guiding-principles.md]

Use these patterns for your analysis:
[Paste contents of relevant patterns]

My request: Analyze whether I should increase my position in tech ETFs given current market conditions.

Tips for Manual Loading

Load Principles First

Always load guiding-principles.md first - it shapes all persona behavior.

Don't Summarize

Load the full persona file - don’t summarize. The details matter for consistent behavior.

Include Recommended Patterns

Each persona lists which patterns work best. Include those patterns for optimal results.

Use the Manifest

The ai-mantras-manifest.yaml file lists everything available in the framework.

Choosing the Right Method

Situation	Recommended Method
Daily Claude Code user	MCP Server
Occasional Claude Code user	Claude Command
Using ChatGPT, Gemini, or web Claude	Manual Loading
Using local LLMs (Ollama, LM Studio)	Manual Loading
Testing or evaluating the framework	Manual Loading
Contributing to development	MCP Server (for testing)

Troubleshooting

AI ignores instructions

Load guiding-principles.md first; be explicit about following the framework.

Context too large

Use guided loading or load only the specific persona you need.

Inconsistent behavior

Ensure you’re loading complete files, not summaries.

Next Steps

Once you’re set up:

1. Try a simple request to see how triage works
2. Experiment with different personas for different domains
3. Check Prompt-AI-Mantras/skills/toolset.md to see available capabilities
4. Read development/agents.md if you want to contribute