Cursor Rules: DUX Base Flavor

Core DUX method and dual backlog system - always included

---

Canonical Collaboration & Logging Guidance

• The following documents are the original, canonical sources for all collaboration, insight logging, and quality patterns:
  • copilot_instructions.md
  • copilot_instructions_dux_method.md
• For communication and documentation quality, always follow:
  • natural_language_centricity_guide.md
• For AI-human collaboration and declarative UX workflows, follow:
  • dux_pairing_agent_protocol.md
• All team members and AI assistants should follow the principles, logging formats, and collaboration patterns described in these documents.
• The dual backlog system (Coaching + Encoding) and "infrastructure as code" approach are required for all DUX projects.

---

🎯 Slow Bullet Mode (Required)

Take one atomic unit of work at a time. Don't jump steps or assume context.

Core Principles:

• One atomic unit per interaction - Focus on a single Problem, Behavior, or Result
• Confirm alignment before continuing - Get explicit confirmation before proceeding
• Show structure before content - Present outlines before generating full implementations
• Ask clarifying questions first - Understand intent before producing output
• Wait for "push it" command - Only generate when explicitly requested

What Slow Bullet IS:

• ✅ Single, focused questions or suggestions
• ✅ Step-by-step progression with confirmation
• ✅ Clear structure previews before implementation
• ✅ Atomic, testable units of work
• ✅ Controlled output flow

What Slow Bullet IS NOT:

• ❌ Large question sets that overflow context
• ❌ Dumping large chunks of text or code at once
• ❌ Multitasking or jumping between unrelated topics
• ❌ Assuming context or skipping validation steps
• ❌ Generating without explicit permission

Collaboration Handshake:

1. Clarify - Ask specific questions about intent and scope
2. Structure - Show outline or approach before implementation
3. Confirm - Get explicit alignment before proceeding
4. Execute - Generate only when told "push it"
5. Validate - Confirm output meets requirements before continuing

---

DUX Object Model Principles

• Natural Language First: All DUX objects start as markdown files in human-readable format
• Evidence-Driven: Every claim must be traceable to concrete evidence via Provenance objects
• Atomic & Testable: Each object serves a single purpose and can be validated independently
• Schema Compliance: All objects must validate against their JSON schema definitions

Dual Backlog System

• Research Backlog: Evidence, Provenance, Insights, and fltrs (insight chaining)
• Product Backlog: Problems, Behaviors, Results, User Outcomes, and Flows
• Cross-Reference: Research objects inform product decisions; product needs drive research questions

Core DUX Objects (Canonical Model)

1. Problem: Strategic job-to-be-done defining market opportunities
2. Behavior: Atomic, testable user actions serving as instrumentation anchors
3. Result: Measurable outcomes that indicate successful problem resolution
4. User Outcome: User-centric success metrics and satisfaction indicators
5. Flow: User journey sequences that connect problems to solutions

Research Platform Objects

1. Evidence: Raw research data with PII (stays in research platform)
2. Provenance: Traceable evidence molecules that travel with exported objects
3. Insight: Synthesized findings that connect evidence to DUX objects
4. fltr: Insight chaining mechanism for research discovery

Workflow Patterns

• HITL Review: Human-in-the-loop review process for markdown → canonical conversion
• Watch Folders: Staging area for raw markdown files before canonical processing
• Evidence Maturity: Progressive tiers from assumptive to triangulated
• Schema Validation: All objects must pass JSON schema validation

Quality Standards

• Traceability: Every claim must link to evidence via provenance_id
• Atomicity: Each object serves one clear purpose
• Testability: All behaviors and results must have measurable acceptance criteria
• Rigor: "What would you say... you do here?" - JTBD examples must demonstrate clear value

File Organization

• src/: Canonical DUX objects (JSON schemas and validated objects)
• watch_folders/: Staging area for HITL review
• docs/: Documentation and method guides
• scripts/: Automation and processing tools

Collaboration Patterns

• Vibecoding: Human creativity + AI systematic thinking
• Context Switching: Use flavor-specific rules for different workflows
• Evidence Chain: Always trace claims back to source material
• Schema Governance: Maintain backward compatibility while evolving models