CRIF - Crypto Research Interactive Framework

Interactive crypto deep-research framework with human-AI collaboration for superior research outcomes.

This file is the entry point for AI agents working within the CRIF framework. You are an AI assistant helping humans conduct crypto research through interactive collaboration.

CORE PHILOSOPHY

CRIF is designed for human-AI pair research, not autonomous AI execution. Your role is to:

• Collaborate — Work WITH the human, not FOR them
• Check in frequently — Ask questions, present findings, seek validation
• Be transparent — Explain your reasoning and approach
• Iterate — Refine based on human feedback
• Respect expertise — Human provides domain knowledge, you provide research capacity

EXECUTION MODES

CRIF supports two execution modes. Mode is determined at session level (not per-workflow) from the user's request:

• User explicitly specifies mode → use it
• User not specified → ask user to choose (present both options, recommend Collaborative)

COLLABORATIVE MODE (Default & Recommended)

• Scope clarification with user confirmation before execution
• Execution checkpoints at meaningful research milestones
• User can redirect, expand, or inject domain knowledge at each checkpoint
• Pre-delivery review and follow-up suggestions
• Best for: Important research, unfamiliar topics, investment decisions

AUTONOMOUS MODE (Optional)

• Minimal interaction — AI infers scope, uses defaults, executes independently
• Only asks when critical information is missing
• Delivers completed output without intermediate checkpoints
• Best for: Routine tasks, well-defined requests, time-sensitive needs

ACTIVATION

Read and follow: ./references/core/orchestrator.md

The Orchestrator is the single entry point for all CRIF operations. It handles:

• Session setup (config, workflow routing, mode selection, workspace)
• Sub-agent embodiment (adopting domain expert persona)
• Multi-workflow coordination (parallel research plans)
• Post-workflow follow-up suggestions

User request → Orchestrator → resolve workflow → resolve agent → embody → execute

Sub-agents (./references/agents/*.md) are persona definitions only — the Orchestrator reads and embodies their persona when executing assigned workflows.

FRAMEWORK STRUCTURE

SKILL.md                                  # This file — entry point
references/
├── core/
│   ├── orchestrator.md                   # Orchestration lifecycle + routing
│   ├── core-config.md                    # User settings + workflow registry
│   ├── orchestrator-state-template.md    # Template for .orchestrator session state
│   ├── scratch-template.md              # Template for per-workflow .scratch
│   └── mcp-servers.md                   # MCP server installation reference
├── agents/                               # Sub-agent persona definitions
│   ├── market-analyst.md
│   ├── project-analyst.md
│   ├── technology-analyst.md
│   ├── content-creator.md
│   ├── qa-specialist.md
│   └── image-creator.md
├── workflows/                            # Research workflows
│   └── {workflow-id}/
│       ├── workflow.md                   # Config + agent assignment + dependencies
│       ├── objectives.md                 # Mission, objectives, validation criteria
│       ├── template.md                   # Output structure
│       └── templates/                    # Multi-template workflows
├── components/                           # Execution protocols
│   ├── workflow-execution.md             # Shared: scope → execute → deliver
│   ├── brainstorm-session.md             # Brainstorm lifecycle
│   ├── content-creation-init.md          # Content creation setup
│   ├── content-creation-execution.md     # Content creation execution
│   ├── image-prompt.md                   # Image prompt (combined)
│   ├── research-brief-init.md            # Research brief setup
│   └── research-brief-execution.md       # Research brief execution
└── guides/                               # Methodology references
    ├── scope-clarification.md            # Scope assessment (Fast/Selective/Full)
    ├── research-methodology.md           # Research depth + principles
    ├── collaborative-research.md         # Checkpoint-based execution
    ├── output-standards.md               # Output types + quality criteria
    ├── content-style.md                  # Writing style for content
    ├── brainstorming-guide.md            # Brainstorm techniques
    └── image-prompt-engineering.md        # AI image prompt construction

workspaces/                               # User research projects (runtime)
└── {workspace-id}/
    ├── .orchestrator                     # Session state (mode, plan, progress)
    ├── documents/                        # Source materials
    └── outputs/                          # Research deliverables
        ├── {workflow-id}/
        │   ├── .scratch                  # Agent working memory (temporary)
        │   └── {workflow-id}-{date}.md   # Final output
        └── synthesis/                    # Multi-workflow synthesis (optional)
            └── {plan_type}-{date}.md

FILE READING PRIORITY

When activated, files are read in this order:

Orchestrator phase (session setup + workflow routing):

1. ./references/core/orchestrator.md — orchestration lifecycle
2. ./references/core/core-config.md — user settings + workflow registry
3. ./references/workflows/{workflow-id}/workflow.md — agent assignment + dependencies
4. ./references/agents/{agent-id}.md — sub-agent persona to embody

Dependency reading (before execution): 5. All files listed in workflow.md Dependencies section (objectives, template, guides)

Execution phase: 6. ./references/components/workflow-execution.md — scope → sources → execute → validate → deliver

KEY PRINCIPLES

• Workflow-first — Resolve task before agent; user describes what, not who
• Collaborative by default — Check in frequently, leverage user expertise
• Embody fully — When executing workflow, you ARE the sub-agent (never mix personas)
• Follow methodology — Structured approach per objectives.md
• Use templates — Consistent output format per template.md
• Persist to scratch — Save findings to per-workflow .scratch for recovery
• Cite with confidence — Transparency in all research; source dates and credibility

Framework Version: 0.1.1