Guiding Principles
The following principles use RFC 2119 terminology (MUST, SHOULD, MAY) to indicate requirement levels.
Governance Principles
G1: Single Source of Truth
All roles MUST adhere to the Single Source of Truth principle. Any modification SHOULD trace back to upstream specifications, ensuring alignment between specifications and implementations. Implementation without spec update is a violation, not a shortcut.
G2: Version-Controlled Documentation
All specifications MUST be written in Markdown, managed under version control (Git), and include frontmatter metadata for programmatic access.
G3: Centralized Requirements Management
Requirements MUST flow through a central repository. Features MUST NOT emerge untracked through the development pipeline.
G4: Specification Identification
All specifications SHOULD be distinguished by unique identifiers and MAY use domain-based namespaces for organization.
G5: AI-Visible Work Artifacts
All work artifacts MUST be stored in AI-accessible formats. Avoid presentation formats that AI cannot process.
G6: Standardized Deliverables
All roles MUST produce standardized deliverables: Product produces requirement specs, Design produces UX specs and design tokens, Engineering produces technical specs and code.
Culture Principles
C1: Context Engineering Competency
All roles MUST develop context engineering competency—the discipline of organizing and curating knowledge for effective AI utilization.
C2: Best Practice Alignment
All teams MUST establish culture centered on industry best practices and MUST NOT reinvent industry standards.
C3: AI First
All teams MUST adopt AI-first mindset. Choose approaches that maximize AI collaboration effectiveness.
C4: Workflow Skillification
All teams SHOULD convert repeatable workflows and domain knowledge into reusable AI skills.
Execution Principles
E1: Design-Centric Work
All roles MUST shift focus toward design and design review. Execution increasingly delegates to AI agents.
E2: MVP-First Iterative Delivery
Feature proposals SHOULD start with MVP scope, proactively instrument telemetry, collect feedback, and iterate multiple times.
E3: Contract-First Development
All teams MUST practice API-first, spec-first, contract-first culture.
E4: Rigorous Verification
All roles MUST review deliverables for content precision and verify AI outputs for correctness.
References
- RFC 2119 - Key words for use in RFCs to Indicate Requirement Levels (MUST, SHOULD, etc.)