Minimal Workflow Example
Overview
This configuration provides the fastest possible SDD workflow with only essential phases:
- Specify - Create requirements
- Implement - Write code
- Done - Complete
Use Cases
- Rapid prototyping
- Personal projects
- Internal tools
- Low-compliance environments
- Proof-of-concept development
- Hackathons
What's Skipped
| Phase | Why Skipped | Risk |
|---|---|---|
| Assessment | Assumes all features suitable for SDD | May waste time on simple fixes |
| Research | Assumes requirements are clear | May build wrong solution |
| Planning | Assumes simple architecture | May encounter tech debt |
| Validation | Relies on manual testing | Quality issues may slip through |
| Operations | Relies on manual deployment | Deployment errors possible |
Workflow Diagram
To Do
↓ /flow:specify
Specified
↓ /flow:implement
In Implementation
↓ manual
Done
Usage
Copy the configuration:
cp docs/examples/workflows/minimal-workflow.yml flowspec_workflow.ymlValidate:
specify workflow validateCreate a task:
backlog task create "Build user profile page"Run the workflow:
/flow:specify # Creates PRD at docs/prd/{feature}.md /flow:implement # Engineers implement, reviewers review backlog task edit task-123 -s Done # Mark complete
Agents Involved
- product-requirements-manager - Creates lightweight PRD
- frontend-engineer - Implements UI
- backend-engineer - Implements API/logic
- frontend-code-reviewer - Reviews UI code
- backend-code-reviewer - Reviews backend code
Customization
Add Basic Testing
Add a validation step before marking Done:
states:
- "To Do"
- "Specified"
- "In Implementation"
- "Tested" # New state
- "Done"
workflows:
validate:
command: "/flow:validate"
agents:
- name: "quality-guardian"
# ...
input_states: ["In Implementation"]
output_state: "Tested"
transitions:
- name: "validate"
from: "In Implementation"
to: "Tested"
via: "validate"
- name: "complete"
from: "Tested" # Changed from "In Implementation"
to: "Done"
Add Planning Phase
For slightly more complex projects:
states:
- "To Do"
- "Specified"
- "Planned" # New state
- "In Implementation"
- "Done"
workflows:
plan:
command: "/flow:plan"
agents:
- name: "software-architect"
# ...
input_states: ["Specified"]
output_state: "Planned"
implement:
input_states: ["Planned"] # Changed from ["Specified"]
# ...
Comparison to Full Workflow
| Aspect | Minimal Workflow | Full SDD Workflow |
|---|---|---|
| Phases | 2 | 7 |
| States | 4 | 9 |
| Agents | 5 | 16 |
| Time to Done | ~1-2 days | ~1-2 weeks |
| Quality Assurance | Manual | Automated + Manual |
| Compliance | None | SOC2/SLSA ready |
| Best For | Prototypes | Production systems |
Migration Path
As projects mature, gradually add phases:
- Start with minimal workflow
- Add Planning when architecture complexity increases
- Add Validation when quality issues appear
- Add Operations when deployment becomes frequent
- Add Research for strategic features
- Add Assessment for portfolio management
Limitations
- No architecture planning (risk of tech debt)
- No security validation (risk of vulnerabilities)
- No compliance support (not suitable for regulated industries)
- No deployment automation (manual deployment overhead)
- No formal QA process (quality depends on code review)
Best Practices
- Still write tests - Even without formal validation phase
- Document architecture decisions - Even without ADRs
- Run security scans - Use pre-commit hooks
- Automate deployment - Use CI/CD even without /flow:operate
- Review regularly - Code review is your quality gate
When to Upgrade
Consider upgrading to a fuller workflow when:
- Project becomes business-critical
- Team size grows beyond 2-3 people
- Compliance requirements emerge
- Deployment frequency increases
- Quality issues increase
- Architecture becomes complex
Related Examples
- Security Audit Workflow - Adds security phase
- Custom Agents Workflow - Organization-specific agents