Structured Development

9-Phase SDLC Workflow

Every project follows a structured workflow with quality gates at each phase. No code until Phase 8. Research, plan, design, then build. AI that follows best practices, not shortcuts.

Click or hover over phases to see descriptions. SPIKE mode is shown as an alternate fast path.

The 5 Questions

Every project is built to answer 5 fundamental questions:

1

SEED - What's the idea?

Capture the initial problem and vision

2

RESEARCH - Who is this for?

Identify target persona and their needs (NEVER "TBD" - always infer)

3

EXPANSION - What are they trying to do?

Understand user intent and existing solutions

4

ANALYSIS & SELECTION - How could they do it?

Evaluate options and choose tech stack (with human approval)

5

DESIGN → TEST → BUILD - What needs to be built?

Design, write tests, then implement the solution

Detailed Phase Breakdown

1

SEED - Capture the Idea

Document the initial problem and vision in a structured format

Deliverables:

  • • Problem statement
  • • Core vision (1-2 paragraphs)
  • • Success criteria
  • • Project classification

Quality Gates:

  • • Problem clearly articulated
  • • Vision is specific, not vague
  • • Scope is defined
  • • UUID assigned

Example:

"Developers waste time context-switching between Claude sessions. Build a plugin system that preserves architectural decisions and domain models across sessions."

2

RESEARCH - Identify the Persona

Who is this for? What are their needs, pain points, and workflows?

Deliverables:

  • • Target persona profile
  • • User needs analysis
  • • Pain points list
  • • Behavioral patterns

Quality Gates:

  • • Persona is specific (never "TBD")
  • • Needs are validated
  • • Pain points are clear
  • • Human confirms persona

Critical Rule:

NEVER use "TBD" for persona. Always infer from context and confirm with human.

3

EXPANSION - Explore Existing Solutions

Research existing tools, patterns, and best practices before building

Deliverables:

  • • Competitive analysis
  • • Existing solution review
  • • Best practices survey
  • • Feature comparison

Quality Gates:

  • • At least 3 alternatives reviewed
  • • Gaps identified
  • • Patterns documented
  • • Citations provided
4

ANALYSIS - Evaluate Options

Compare approaches and define success metrics

Deliverables:

  • • Option matrix
  • • Trade-off analysis
  • • Success metrics
  • • Risk assessment

Quality Gates:

  • • Options clearly defined
  • • Trade-offs explicit
  • • Metrics measurable
  • • Risks documented
5

SELECTION - Choose Tech Stack

Select technologies with human approval (protected by guardrails)

Deliverables:

  • • Tech stack proposal
  • • Architecture approach
  • • Dependency list
  • • Rationale document

Quality Gates:

  • • Guardrails checked
  • • Human approval obtained
  • • Dependencies justified
  • • Rationale documented

⚠️ Human Approval Required

AI cannot proceed to Phase 6 without explicit human approval of tech choices.

6

DESIGN - Create Specs

Design mockups, architecture diagrams, and domain models

Deliverables:

  • • UI mockups (if applicable)
  • • Architecture diagrams
  • • Domain models
  • • API specifications

Quality Gates:

  • • Design tokens used
  • • Domain models atomic
  • • Diagrams complete
  • • Specs reviewed
7

TEST - Write Tests First (TDD)

Write tests BEFORE implementation. No code in this phase.

Deliverables:

  • • Unit tests
  • • Integration tests
  • • BDD scenarios (Cucumber)
  • • Test data fixtures

Quality Gates:

  • • All features have tests
  • • Edge cases covered
  • • Tests are runnable
  • • No implementation code

🚫 NO CODE in Phase 7

Only test files. Implementation happens in Phase 8.

8

IMPLEMENTATION - Build the Solution

Finally write code! Follow design specs and make tests pass.

Deliverables:

  • • Working implementation
  • • Passing tests
  • • API documentation
  • • Deployment scripts

Quality Gates:

  • • All tests passing
  • • Code follows standards
  • • Types enforced
  • • Docs updated

✅ First Phase with Code

Phases 1-7 are planning. Phase 8 is where implementation begins.

9

REFINEMENT - Polish & Optimize

Performance tuning, error handling, monitoring, production prep

Deliverables:

  • • Performance optimizations
  • • Error handling
  • • Monitoring setup
  • • Production deployment

Quality Gates:

  • • Performance benchmarks met
  • • Error boundaries present
  • • Logs configured
  • • Smoke tests passing

Special Mode: SPIKE

Need to move fast? SPIKE mode is for throwaway prototypes

When to Use SPIKE:

  • • Proof-of-concept experiments
  • • Quick technology evaluation
  • • Throwaway demos
  • • Learning new frameworks
  • • Time-boxed research (max 3 days)

What SPIKE Skips:

  • × Phase 2: Research (persona analysis)
  • × Phase 3: Expansion (competitive review)
  • × Phase 4: Analysis (option evaluation)
  • × Phase 5: Selection (human approval)
  • × Phase 6: Design (specs and mockups)
  • × Phase 7: Test (TDD)

Warning: SPIKE code is disposable

SPIKE prototypes are for learning and exploration only. If successful, restart from Phase 1 to build production version.

Why 9 Phases?

Prevents Over-Engineering

Research existing solutions before building. Don't reinvent wheels.

Validates Assumptions Early

Persona and analysis phases catch bad ideas before writing code.

Enforces TDD

Tests written before implementation ensure behavior-driven design.

Preserves Context

Each phase documents decisions for future reference.

Reduces Rewrites

Design phase catches architecture issues before coding.

Team Alignment

Structured phases make it easy for teams to collaborate.

Build Better Software with Structured Workflow

Stop jumping straight to code. Let ProtoFlow's 9-phase SDLC guide you from idea to production with quality gates at every step.

Get Started Back to Home →