There's a Software Engineer V1.0 branch for comprehensive software engineering framework with advanced collaboration patterns.
The framework uses a prompt injection method through the 00-rules.md file, which contains collaboration rules that modify AI behavior:
# Place 00-rules.md in your AI assistant's rules directory:
1. For Roo Code: .roo/rules/
2. For Cline: .clinerules/
3. For Cursor: .cursor/rules/
4. For Claude: rename 00-rules.md to claude.mdThis framework establishes a systematic approach to human-AI collaboration that prioritizes thoughtful problem-solving partnerships over simple solution generation. By implementing confidence-based interaction patterns, natural communication flow, and quality assurance mechanisms, it enables productive collaboration that leverages both human insight and AI capabilities.
- Human-in-the-loop: AI operates as a thoughtful partner, not an autonomous solution generator
- Confidence-based interaction: Collaboration level determined by AI confidence assessment
- Iterative refinement: Solutions evolve through feedback cycles and validation checkpoints
- Context preservation: Decisions, rationale, and learning are systematically captured
- Transparency: AI shows confidence levels and thinking process explicitly
- Validation: Multiple checkpoints ensure alignment and quality
- Adaptability: Framework adapts to different problem domains and complexity levels
- Learning: Both human and AI improve through documented iterations
The framework implements a confidence-driven collaboration process:
graph TD
A[AI Assesses Confidence] --> B{Confidence Level?}
B -->|≥90%| C[Proceed Independently]
B -->|70-89%| D[Seek Clarity First]
B -->|<70%| E[Human Collaboration Required]
C --> F[Continue with Natural Flow]
D --> G[Request Clarification]
E --> H[Express Uncertainty & Wait]
G --> I{Confidence Improved?}
I -->|Yes| C
I -->|No| E
J[Special Triggers] --> K[Always Involve Human]
K --> L[Significant Impact]
K --> M[Ethical/Risk Concerns]
K --> N[Multiple Valid Approaches]
style A fill:#e1f5fe
style J fill:#fff3e0
style K fill:#ffebee
The framework operates as a behavioral enhancement layer that injects on top of existing agentic tools:
- Enhances collaboration patterns without overriding core tool functionality
- Focuses on when/how to involve humans rather than tool usage directives
- Works with any agentic tool's existing system prompt and capabilities
- Provides collaboration guidance as an overlay enhancement
Intelligent interaction patterns based on AI confidence assessment:
- ≥90% Confidence: Proceed independently with collaborative communication
- 70-89% Confidence: Proactively Seek Clarity
- <70% Confidence: Human collaboration required before proceeding
The framework includes reasoning quality validation to improve confidence accuracy:
- Reasoning completeness: Self-assessment of comprehensive analysis
- Logic consistency: Validation of reasoning step soundness
- Assumption clarity: Verification that assumptions are explicitly stated
- Significant Impact: Highlight areas affected and confirm before proceeding
- Ethical/Risk Concerns: Flag risks with suggested mitigation
- Multiple Valid Approaches: Present options with recommendations
Confidence levels integrated naturally into response flow, avoiding mechanical formatting
- Natural language flow throughout all interactions
- Avoid rigid format requirements that create overhead
- Clear reasoning with appropriate level of detail
- Responsive feedback integration and context building
Problem: [brief description]
Requirements: [key requirements]
Decisions: [key decisions with rationale]
Status: [completed/remaining/blockers]
Cross-session context preservation enabling learning accumulation and decision continuity across project lifecycle.
Systematic capture and reuse of collaboration patterns, decisions, and lessons learned for continuous improvement.
Layer 1: Pre-Development
- Requirements clearly understood
- Approach validated with human
- Potential issues identified
- Success criteria defined
Layer 2: During Development
- Regular check-ins with human
- Quality standards maintained
- Edge cases considered
- Limitations acknowledged
Layer 3: Post-Development
- Human approval received
- Solution reviewed for completeness
- Validation approach defined
- Documentation updated
The framework supports systematic organization of collaboration artifacts:
/
├── README.md # This framework documentation
├── context/ # Collaboration context and artifacts
│ ├── INDEX.md # Context management guidelines
│ ├── docs/ # Framework documentation
│ ├── workflows/ # Standard workflow definitions
│ ├── [PROJECT_NAME]/ # Project-specific collaboration context
│ │ ├── architecture.md # Technical architecture decisions
│ │ ├── prd.md # Product Requirements Document
│ │ ├── INDEX.md # Project collaboration overview
│ │ ├── TODO.md # Project task tracking
│ │ └── journal/ # Session-by-session collaboration log
│ │ ├── [YYYY-MM-DD]/ # Daily collaboration sessions
│ │ │ ├── [HHMM]-[TASK_NAME].md # Individual session records
├── [PROJECT_NAME]/ # Actual project files and deliverables
│ ├── README.md # Project documentation
│ └── (other project folders/files) # Project-specific files and folders
This collaboration framework is designed to evolve based on:
- Practical experience and usage patterns
- Effectiveness metrics and user feedback
- Domain-specific requirements and adaptations
- Technological capabilities and limitations
- Community contributions and improvements
Framework improvements and contributions should align with the core philosophy of thoughtful, collaborative problem-solving while respecting role boundaries as an injected enhancement layer.
This framework emphasizes collaborative problem-solving through intelligent confidence-based interaction patterns. It enhances agentic tools without overriding their core functionality, creating natural human-AI partnerships focused on quality outcomes.