CIRF - Crypto Interactive Research Framework

A structured prompt engineering framework for conducting comprehensive crypto research with AI assistance.

Author: Kudō

---

What is CIRF?

CIRF (Crypto Interactive Research Framework) is a specialized framework designed for crypto market research that emphasizes human-AI interaction for optimal control and output quality.

The Problem with Autonomous AI Research

Traditional AI research approaches let models work freely and autonomously - you give a prompt, AI disappears for hours, then returns with a completed report. This creates several issues:

• ❌ No visibility into the research process
• ❌ No ability to course-correct during execution
• ❌ Outputs may miss your specific needs or priorities
• ❌ Black-box process with limited quality control

The CIRF Approach: Interactive Collaboration

Instead of autonomous execution, CIRF transforms research into a collaborative team session between you and AI:

• ✅ Continuous control - Guide AI through each research phase
• ✅ Real-time adjustments - Refine focus areas as insights emerge
• ✅ Interactive validation - Review findings before moving forward
• ✅ Quality assurance - Ensure outputs match your expectations
• ✅ Transparency - Understand AI's reasoning at each step

Think of it as pair programming for research - you're the domain expert directing, AI is the research assistant executing.

CIRF is a prompt engineering framework written entirely in natural language (YAML + Markdown) with zero lines of code. This means you can easily read, understand, and customize any part of the framework to fit your needs.

Flexibility: Collaborative or Autonomous

While CIRF is designed for interactive collaboration (recommended), it also supports autonomous mode when appropriate:

• Collaborative mode (Default) - Step-by-step guidance, review at each phase
• Autonomous mode (Optional) - Full workflow execution, minimal intervention

Best practice: Start collaborative to build understanding, then use autonomous for repetitive tasks.

---

How It Works

CIRF enables iterative human-AI collaboration through continuous interaction:

┌─────────────────────────────────────────────┐
│         INTERACTIVE RESEARCH LOOP            │
└─────────────────────────────────────────────┘

  YOU DIRECT  →  AI RESEARCHES  →  AI CHECKS IN
      ↑                                  ↓
      │                                  ↓
      └──────  YOU REVIEW & REFINE  ←───┘
                       │
                       ↓
                 ITERATE UNTIL
                  SATISFIED
                       │
                       ↓
                 FINAL OUTPUT

The Loop:

1. You direct - Define goal, set priorities, provide context
2. AI researches - Gathers data, conducts analysis following framework methodology
3. AI checks in - Presents findings, asks clarifying questions
4. You review & refine - Validate, redirect, add expertise
5. Iterate - Loop continues through research phases until satisfied
6. Final output - AI generates structured report following template

This continuous feedback loop ensures research stays aligned with your goals and incorporates your expertise at every critical decision point.

---

Requirements

AI Assistant

This framework works with AI CLI tools that have file read/write capabilities:

• ✅ Claude Code (Recommended)
• ✅ OpenClaw
• ✅ Codex CLI
• ✅ Any AI assistant with file system access

Setup

git clone https://github.com/kudodefi/cirf.git
cd cirf

That's it. No installation, no dependencies.

---

Quick Start

Method 1: Agent File Path (Recommended)

Tag the agent file directly to activate:

You: @framework/agents/research-analyst.yaml
     I need competitive analysis for Ethereum vs Bitcoin.
     Use collaborative mode.

AI: [Reads and embodies research analyst persona]
    Running competitive-analysis workflow in collaborative mode...

Why recommended:

• ✅ Direct and explicit
• ✅ AI knows exactly which file to read
• ✅ Works reliably across all AI tools

Available agent files:

• @framework/agents/research-analyst.yaml - Market & fundamentals
• @framework/agents/technology-analyst.yaml - Technical architecture
• @framework/agents/content-creator.yaml - Content transformation
• @framework/agents/qa-specialist.yaml - Quality assurance

Method 2: Agent Tag Shorthand

Use agent name as shorthand:

You: @Research-Analyst - Analyze Bitcoin's market position.
     Collaborative mode.

AI: [Interprets and reads framework/agents/research-analyst.yaml]
    Starting analysis...

Method 3: Natural Language

Simply describe what you want:

You: I want to analyze Ethereum's current market position.
     Help me understand its competitive landscape.

AI: I'll activate as Research Analyst and suggest appropriate workflows.

Method 4: Orchestrator Mode

For complex multi-workflow research:

You: Read @SKILL.md and act as orchestrator.
     I want comprehensive analysis of Ethereum for investment decision.

AI: [Reads SKILL.md orchestration protocol]
    [Proposes multi-workflow research plan]
    [Executes with your approval at each phase]

---

Framework Structure

cirf/
├── README.md                  # This file (for humans)
├── SKILL.md                   # Orchestration instructions (for AI)
│
├── framework/
│   ├── core-config.yaml       # Framework configuration
│   ├── _workspace.yaml        # Workspace template
│   │
│   ├── agents/                # 4 Expert personas
│   │   ├── research-analyst.yaml
│   │   ├── technology-analyst.yaml
│   │   ├── content-creator.yaml
│   │   └── qa-specialist.yaml
│   │
│   ├── workflows/             # 17 Research workflows
│   │   └── {workflow-id}/
│   │       ├── workflow.yaml
│   │       ├── objectives.md
│   │       └── template.md
│   │
│   ├── components/            # Shared execution protocols
│   └── guides/                # Research methodologies
│
└── workspaces/                # Your research projects (auto-created)
    └── {project-id}/
        ├── workspace.yaml
        ├── documents/
        └── outputs/

---

Expert Agents

Each agent is a persona definition that AI embodies when activated.

🔬 Research Analyst

Expertise: Market intelligence, project fundamentals, investment synthesis Use for: Market analysis, project evaluation, competitive research, investment thesis Workflows: All research workflows

Activate:

@framework/agents/research-analyst.yaml

⚙️ Technology Analyst

Expertise: Architecture assessment, security analysis, technical evaluation Use for: Smart contract review, protocol architecture, technical due diligence Workflows: technology-analysis

Activate:

@framework/agents/technology-analyst.yaml

✍️ Content Creator

Expertise: Research-to-content transformation, multi-platform optimization Use for: Converting research into X threads, blog articles, YouTube scripts Workflows: create-content

Activate:

@framework/agents/content-creator.yaml

✓ QA Specialist

Expertise: Quality validation, critical review, bias detection Use for: Reviewing research outputs, challenging assumptions, finding gaps Workflows: qa-review

Activate:

@framework/agents/qa-specialist.yaml

---

Workflows

Each workflow is a structured research process with defined methodology and output template.

Setup & Planning

Market Intelligence

Project Fundamentals

Technical & Quality

Content & Flexible

---

Workspace & Output Management

Each research project gets its own workspace for organization and isolation.

Workspace Structure

workspaces/{project-id}/
├── workspace.yaml          # Project metadata & configuration
├── documents/              # Source materials, references
└── outputs/                # Generated research deliverables
    └── {workflow-id}/
        └── {workflow-id}-{date}.md

How It Works

AI automatically manages workspaces:

You: Analyze Ethereum for investment.

AI: Creating workspace 'ethereum'...
    ✅ Workspace ready: workspaces/ethereum/
    Starting research...

Outputs are auto-saved with version control:

outputs/sector-overview/
├── sector-overview-2025-02-09.md  # Latest
├── sector-overview-2025-02-01.md  # Previous
└── sector-overview-2025-01-15.md  # Initial

Context preserved across sessions:

Session 1: sector-overview
Session 2: competitive-analysis (builds on session 1 context)
Session 3: synthesis (consolidates all previous work)

---

Best Practices

1. Prefer Collaborative Mode

Why collaborative is recommended:

• ✅ Quality control at each phase
• ✅ Inject domain expertise when needed
• ✅ Course-correct as insights emerge
• ✅ Ensure outputs match expectations

When to use autonomous:

• ✅ Repetitive analysis with clear requirements
• ✅ After calibrating AI's approach collaboratively
• ✅ Time-sensitive updates using established framework

2. Build Research in Phases

Don't try to do everything at once:

Phase 1: sector-overview (market context)
Phase 2: project-snapshot (subject overview)
Phase 3: Specialized deep-dives (based on Phase 1-2 findings)
Phase 4: Synthesis (investment thesis or recommendation)

3. Use Research Briefs for Complex Projects

Define clear objectives upfront:

You: Create research brief for Ethereum analysis.

AI: [Helps structure objectives, scope, approach]
    [Suggests workflow sequence]

4. Leverage AI for Research, You for Judgment

AI excels at: Data gathering, synthesis, pattern identification You excel at: Strategic direction, domain expertise, investment judgment

You: Analyze competitive moat. I believe network effects matter more than tech specs.

AI: Understood. Weighting analysis: High priority on network effects, switching costs.
    [Executes with your strategic framing]

5. Iterate on Outputs

Don't settle for first draft:

AI: Report complete.

You: Competitive landscape needs more depth. Add emerging players.

AI: [Refines section]
    Better?

You: Good. Now add risk factors for each incumbent.

AI: [Final version saved]

6. Use QA Review for Critical Decisions

Validate important research:

You: @framework/agents/qa-specialist.yaml
     Review my Ethereum investment analysis.

AI: ⚠️ Potential bias: Heavy weight on developer ecosystem, value accrual link unclear.
    📊 Gap: Missing regulatory risk for DeFi platforms.
    ❓ Logic: Bull case assumes L2 success but doesn't model fragmentation risk.

    Recommend addressing before final decision.

7. Start Broad, Then Go Deep

Recommended sequence:

1. Sector-overview (context)
2. Project-snapshot (overview)
3. Specialized workflows (deep dives)
4. Synthesis (conclusions)

8. Use Autonomous for Monitoring

After initial collaborative research, use autonomous for updates:

Initial: Comprehensive Bitcoin analysis (collaborative, multi-session)
3 months later: Update with Q1 2025 data (autonomous, same framework)

---

Troubleshooting

AI doesn't understand framework

Solution: Explicitly tell AI to read framework files

You: Read @SKILL.md first, then help me with research.

Outputs not following template

Solution: Remind AI to follow template

You: Execute sector-overview. Follow template.md exactly.

Research lacks depth

Solution: Specify research depth

You: Run product-analysis with deep research depth.
     Follow methodology in framework/guides/research-methodology.md.

---

Contributing

Found issues or have suggestions? Open an issue or submit a pull request.

---

License

MIT License - See LICENSE file for details.

---

Credits

Created by: Kudō Framework Version: 1.0.0 Last Updated: 2025-02-09

---

Ready to start researching?

You: @research-analyst.yaml
     Help me research [your topic]