SDLC Workflow Automation
I automated most of my Android SDLC with 4 Claude Code commands.
Not code generation only - the full lifecycle: story refinement, design, implementation, and verification.
It started with Spec-Driven Development (thanks to @Tabish Bidiwale's OpenSpec), but I needed more. So I built a layer on top, that handles the entire user story lifecycle.
After 2 months of iteration across multiple Android projects, it boiled down to 4 commands:
* /sdlc_open_story - picks the next story, creates a branch, refines it with a full product-level analysis
* /sdlc_propose_change - generates design artifacts: proposal, specs, and BDD-style tasks with GIVEN/WHEN/THEN acceptance criteria
* /sdlc_implement_change - executes via sub-agent orchestration (fresh context, cheaper models), with the parent agent verifying each step
* /sdlc_verify_story - end-to-end gate: unit tests, on-device tests, security review, acceptance criteria check, documentation sync
Each command feeds into the next - but the human reviews and approves before moving forward. That's the key. The dev stays in control, catches bad decisions early, and those corrections feed back into the agent's guidelines.
Guidelines (Android best practices, git conventions, project-specific rules) and MCP-connected story tracking make it all work together. Every story generates a section in an HTML report for full traceability.
Both repos are public:
* Framework: https://github.com/guidovezzoni/SDLC
* Test project (clone it, open in Claude Code, run `/sdlc_open_story`): https://github.com/guidovezzoni/venice
What's your experience with AI-assisted development workflows? Still prompting ad-hoc or moving toward something structured?