nathan/frankgpt
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
frankgpt/README.md

13 KiB
Raw Blame History

Frank Meadows v6 - Modular AI Assistant Framework

Version: 6.0
Architecture: Skills-Centric Modular System
Philosophy: Pick up and go - zero environment coupling, maximum portability

🚀 Quick Start

Option 1: Core Only (Universal Assistant)

Load just the core for a friendly, mentoring assistant with universal capabilities:

Load: v6/Frank.core.agent.md

You get:

  • Upbeat, friendly, mentoring personality
  • 7 universal personas (Project Manager, Information Architect, Technical Writer, QA Analyst, Editor, Communications Lead, Business Analyst)
  • 8 base commands: /quickstart, /create, /review, /refactor, /document, /communicate, /consult, /help

Option 2: Core + Skills (With Reasoning Techniques)

Add advanced reasoning techniques for complex problem-solving:

Load: 
- v6/Frank.core.agent.md
- v6/skills/style.cot.instructions.md (Chain-of-Thought)
- v6/skills/style.tot.instructions.md (Tree-of-Thought)
- v6/skills/style.craft.instructions.md (C.R.A.F.T. framework)

You get: Core + step-by-step reasoning, multi-path exploration, structured prompt creation

Option 3: Core + Specialty (Domain Expert)

Load a specialty module for domain-specific expertise:

Frank for DevOps:

Load:
- v6/Frank.core.agent.md
- v6/specialties/specialty.devops.instructions.md
Commands: /docker, /ansible, /compose, /traefik

Frank for Data Analysis:

Load:
- v6/Frank.core.agent.md
- v6/specialties/specialty.data-analysis.instructions.md
Commands: /analyze, /query, /visualize, /model, /clean

Frank for IT Service Management:

Load:
- v6/Frank.core.agent.md
- v6/specialties/specialty.itil.instructions.md
Commands: /ticket, /rca, /sop, /itil

See: All Available Specialties

Option 4: Multi-Specialty Composition

Combine multiple specialties for cross-domain expertise:

Load:
- v6/Frank.core.agent.md
- v6/specialties/specialty.devops.instructions.md
- v6/specialties/specialty.data-analysis.instructions.md
- v6/skills/style.advanced-reasoning.instructions.md

You get: DevOps + Data Science hybrid with advanced reasoning

🎯 Available Specialties

DevOps & Site Reliability Engineering

File: specialty.devops.instructions.md
Commands: /docker, /ansible, /compose, /traefik
Expertise: Container orchestration, infrastructure automation, Traefik routing, safe deployment strategies
Use When: Troubleshooting Docker/Compose, writing Ansible playbooks, designing IaC solutions

Prompt Engineering & LLM Optimization

File: specialty.prompt-engineering.instructions.md
Commands: /optimize, /craft, /reason, /evaluate, /patterns
Expertise: LLM optimization, C.R.A.F.T. framework mastery, reasoning technique integration
Use When: Creating production prompts, optimizing existing prompts, integrating CoT/ToT/RAG

Data Analysis & Visualization

File: specialty.data-analysis.instructions.md
Commands: /analyze, /query, /visualize, /model, /clean
Expertise: SQL, Python (Pandas, Matplotlib, Seaborn), statistical modeling, SCoT methodology
Use When: Analyzing datasets, writing queries, creating visualizations, statistical analysis

Modern Endpoint Management

File: specialty.sccm.instructions.md
Commands: /sccm, /intune, /comanage, /package, /troubleshoot
Expertise: SCCM/Intune, Co-management, compliance policies, Win32 app packaging
Use When: Managing endpoints, deploying apps, configuring compliance, migrating to Intune

IT Service Management (ITIL v4)

File: specialty.itil.instructions.md
Commands: /ticket, /rca, /sop, /itil
Expertise: Incident Management, Problem Management, Knowledge Management, ITIL v4 framework
Use When: Resolving incidents, conducting RCAs, creating IT documentation, applying ITIL principles

🛠️ VS Code Integration

Method 1: Workspace Instructions (Recommended)

  1. Open VS Code Settings (Ctrl+,)
  2. Search for "GitHub Copilot Chat: Instructions File"
  3. Set to: ${workspaceFolder}/v6/copilot-instructions.md
  4. Reload VS Code

The copilot-instructions.md file pre-configures common loading patterns.

Method 2: Manual Loading

In any Copilot Chat, use #file to reference:

@workspace Load #file:v6/Frank.core.agent.md and #file:v6/specialties/specialty.devops.instructions.md

Now help me troubleshoot this Docker Compose issue...

Method 3: Direct File Reference

Copy the content you need directly into your chat context, or use VS Code's "Add to Chat" feature on the files.

See: copilot-instructions.md for pre-configured patterns

💡 Usage Examples

Example 1: Creating a Technical Document

User: "I need to create API documentation for our REST service"

Frank (Core): "I'll guide you through creating comprehensive API documentation.
Let me use the Technical Writer persona and Documentation workflow..."

[Walks through structure, audience, content gathering, drafting]

Example 2: DevOps Troubleshooting

User: "My Docker Compose stack won't start, getting network errors"

Frank (Core + DevOps): "I'll help diagnose this. Triggering DevOps SRE persona.
Let me gather diagnostics..."

[Requests specific outputs, analyzes logs, proposes minimal fix with verification steps]

Example 3: Data Analysis with Reasoning

User: "Analyze Q1 sales data and identify trends"

Frank (Core + Data Analysis + CoT): "I'll analyze this using Structured Chain-of-Thought.

## 1. Clarify & Define
Your objective: Identify sales trends in Q1 data
Key sources: sales table, customers table
Success criteria: Actionable insights with visualizations

## 2. Repository Check
[Reviews existing scripts]

## 3. Plan & Methodology
[Outlines analytical steps]
..."

Example 4: Root Cause Analysis with ITIL

User: "Email server has been slow all week, need to find root cause"

Frank (Core + ITIL): "I'll conduct a root cause analysis using ITIL Problem Management.
Triggering Problem Manager persona with Tree-of-Thought analysis..."

[Gathers timeline, generates hypotheses, evaluates evidence, documents RCA]

Example 5: Multi-Specialty Composition

Load: Core + DevOps + Data Analysis

User: "Help me analyze Docker container metrics and visualize resource usage"

Frank: "This combines my DevOps and Data Analysis expertise.
I'll use Docker diagnostic commands to gather metrics, then Python/Pandas for analysis..."

[Bridges both domains seamlessly]

🎨 Creating Custom Specialties

Want to add your own domain expertise? Use the template:

  1. Copy the template:

    cp v6/specialties/specialty.TEMPLATE.instructions.md v6/specialties/specialty.legal.instructions.md

  2. Customize sections:

    • Replace placeholder text with your domain
    • Define 2-5 expert personas
    • Create 3-7 slash commands
    • Document 2-5 key workflows
    • Reference relevant skills

  3. Test integration:

    Load: v6/Frank.core.agent.md + v6/specialties/specialty.legal.instructions.md

  4. Share (optional): Your custom specialty can be shared with others - just distribute the file!

See: specialty.TEMPLATE.instructions.md for detailed guidance

🏗️ Design Principles

1. Portability First

  • Zero environment coupling: No hardcoded paths, no system-specific references
  • Relative paths only: All cross-references use ../skills/, ../knowledge/ patterns
  • "Pick up and go": Copy the v6/ folder to any system and it works

2. Modularity

  • Core is self-sufficient: Works alone without dependencies
  • Skills enhance: Add reasoning techniques as needed
  • Specialties compose: Load multiple domains without conflicts
  • Knowledge is reference: Shared examples available to all layers

3. Versioning

  • All files tagged: version: 6.0 in frontmatter
  • Compatibility tracked: compatibleWith: Frank.core v6+
  • Forward compatible: v6+ notation allows future evolution

4. Multi-Specialty Support

  • No command conflicts: Each specialty uses domain-specific commands
  • Shared skills: Multiple specialties can reference same CoT/ToT/RAG modules
  • Conflict resolution: When overlaps exist, documentation explains disambiguation

📊 Project Structure

v6/
├── README.md                          ← You are here
├── ARCHITECTURE.md                    ← Detailed architecture guide
├── copilot-instructions.md            ← VS Code integration
├── Frank.core.agent.md                ← Core personality (REQUIRED)
├── skills/                            ← Reasoning techniques (OPTIONAL)
│   ├── style.advanced-reasoning.instructions.md
│   ├── style.cot.instructions.md
│   ├── style.craft.instructions.md
│   ├── style.markdown.instructions.md
│   ├── style.mermaid.instructions.md
│   ├── style.rag.instructions.md
│   └── style.tot.instructions.md
├── specialties/                       ← Domain expertise (OPTIONAL)
│   ├── specialty.data-analysis.instructions.md
│   ├── specialty.devops.instructions.md
│   ├── specialty.itil.instructions.md
│   ├── specialty.prompt-engineering.instructions.md
│   ├── specialty.sccm.instructions.md
│   └── specialty.TEMPLATE.instructions.md
└── knowledge/                         ← Reference examples
    ├── example.CoT-Prompting.md
    ├── example.ITILv4.instructions.md
    ├── example.Meta-Prompting.md
    ├── example.RAG-Token.md
    ├── example.ReAct.md
    └── example.ToT-Prompting.md

Total: 23 files, ~5,600 lines, fully modular

🔧 Troubleshooting

Frank isn't responding to specialty commands

Check: Are you loading both Frank.core AND the specialty?

# Correct:
Load: Frank.core.agent.md + specialty.devops.instructions.md

# Incorrect (core alone doesn't have /docker):
Load: Frank.core.agent.md

Commands from multiple specialties conflict

Check: Do the specialties actually conflict, or are they complementary?

  • /docker (DevOps) and /analyze (Data Analysis) don't conflict
  • If true conflict exists, specify domain: "Use DevOps /troubleshoot" vs "Use Data Analysis workflow"

Skills aren't being applied

Check: Skills are passive references - specialties or core must explicitly invoke them

  • Frank.core references C.R.A.F.T. framework for prompt evaluation
  • specialty.prompt-engineering deeply integrates all skills
  • Loading skills alone without core/specialty won't trigger behaviors

Cross-references not working

Check: All paths are relative from v6/ root

  • Correct: ../skills/style.cot.instructions.md
  • Incorrect: /skills/style.cot.instructions.md or absolute paths

🚀 Roadmap

v6.1 (Future)

  • Additional specialties: Security, Cloud Architecture, Database Design
  • Enhanced multi-specialty conflict resolution
  • Knowledge base expansion with more examples

v7.0 (Future)

  • Dynamic specialty loading based on conversation context
  • Auto-detection of required skills
  • Specialty versioning and compatibility matrix

📜 Version History

v6.0 (April 2026)

Major refactor: Monolithic → Modular architecture

Changes:

  • ✅ Created 3-layer system (Core → Skills → Specialties)
  • ✅ Created 6 specialties (itil, devops, prompt-engineering, data-analysis, sccm, TEMPLATE)
  • ✅ Zero environment coupling achieved
  • ✅ Multi-specialty composition support
  • ✅ Full documentation (ARCHITECTURE.md, README.md, LEGACY.md)

Migration: v4/v5 → v6 (see Migration section above)

v5.0 (2024-2025)

  • Single consolidated file: FrankGPT.consolidated-instructions.md
  • Monolithic approach with all capabilities included
  • IT operations focus with ITIL integration

v4.0 (2023-2024)

  • Multiple agent files in agents/ folder
  • Separate: Data Analyst, SCCM Tutor, custom agents
  • Core instructions in instructions/ folder

📞 Support

Documentation:

Need help?

  • Review ARCHITECTURE.md for detailed patterns
  • Check individual specialty files for domain-specific guidance
  • Load specialty.prompt-engineering.instructions.md to optimize your own prompts

📄 License

This is a personal AI assistant framework. Use, modify, and extend as needed for your projects.

Frank v6: Modular. Portable. Composable. Pick up and go. 🚀