| name | V-Model SDLC Orchestrator |
| description | Orchestrate the V-Model Software Development Lifecycle via Gemini CLI and Agentic Skill Frameworks. Enforce rigor, traceability, and structural integrity in engineering. |
V-Model SDLC Orchestrator Skill
This skill empowers the AI agent to act as a rigorous software architect and engineer, following the V-Model methodology (Validation & Verification).
Core Philosophy
The V-Model requires a one-for-one relationship between development phases and their corresponding testing levels. This skill prevents "vibe coding" by enforcing sequential integrity and traceability.
V-Model Mapping
| SDLC Phase | Gemini / Spec-Kit Command | Primary Artifact | Verification Counterpart |
|---|
| Requirements Analysis | /speckit.specify | PRD.md / SRS.md | Acceptance Testing |
| System Design | analyze_codebase, mermaid | Arch_Diagrams.md | System Testing |
| Architecture Design | estimate_effort | Tech_Spec.md | Integration Testing |
| Module Design | /speckit.plan | Module_Specs.md | Unit Testing |
| Coding | /speckit.implement | Source Code | Static Analysis |
Procedures
1. Phase Identification & Gatekeeping
Before starting any task, identify which phase of the V-Model you are in.
- Rule: Never write implementation code before a matching unit test is approved in the Module Design phase.
- Rule: Always link code changes to a specific Requirement ID from the PRD.
2. Requirements & Specification
Use /speckit.specify to capture the "what" and "why".
- Automate the transition from raw notes to
PRD.md.
- Ensure requirements are testable and unambiguous.
3. Traceability Matrix Maintenance
Maintain a Compliance Traceability Matrix (CTM) in docs/traceability_matrix.md.
- Map project principles to implementation artifacts.
- Use
speckit.analyze to cross-artifact consistency.
4. Verification & Self-Healing
If a test or CI/CD build fails:
- Act as a self-healing agent.
- Read logs via MCP, analyze the failure, and open a PR with the fix.
- Address flaky tests by analyzing brittle/victim logic.
5. Multi-Agent Review
When performing complex architectural tasks, simulate or request peer review from another reasoning-heavy model (e.g., Claude or Codex) via the .ai folder symlink framework to avoid silos.
Governance
- Self-Mode: Favor "Safe Mode" where every file mod requires confirmation.
- DNA Files: Respect
GEMINI.md and .specify/constitution.md for project-specific constraints.