Built by Superml.dev & superml.org by crazyaiml
Smart SDLC is an AI-driven development framework delivered as skill files your AI assistant reads and follows — no runtime, no plugins, no external services required.
| Document | What It Covers |
|---|---|
| Personas | The 8 AI personas, their roles, skills, and how the persona-lock system works |
| General & API Workflow | End-to-end workflow for new product, feature, or API projects |
| Modernization Workflow | Full legacy analysis → migration planning → delivery workflow |
| Greenfield Workflow | Workflow for net-new projects starting from a blank canvas |
| Skills Reference | Every skill: what it does, which persona owns it, how to activate it |
| Integrations | JIRA, Confluence, GitHub, GitLab, Azure DevOps setup and usage |
A skill is a markdown file (SKILL.md) containing structured instructions. When your AI reads a skill file, it adopts that workflow — no code execution required. Skills have two forms:
- Agent skills (e.g.
agent-pm/SKILL.md) — define a persona that handles a whole domain - Workflow skills (e.g.
create-prd/SKILL.md) — drive a specific multi-step task
A persona is a named AI role with a defined skill set. Each persona has:
- A primary agent skill that loads on activation
- A curated set of workflow skills it can access
- A Persona Guard on every skill — prevents accidentally using skills outside your role
Eight personas ship with Smart SDLC: Scout, Product/BA, Architect, Developer, Team Lead, QA, Release, and Modernization Lead. Default names are overridden during init.
Every generated skill (in .github/skills/) includes a Persona Guard section. When activated, the AI:
- Reads
_superml/persona.ymlsilently - If the user's
primary_personamatches the skill's allowed personas → proceeds - If not → stops and redirects:
/sml-meeting— bring the right persona into the session/sml-help— see skills for your own role
The meeting command generates a multi-persona context prompt that brings two or more personas into one AI session. This is the correct way to do cross-persona work (e.g. Architect reviewing BA's PRD, or Developer asking BA to clarify acceptance criteria).
_superml/config.yml tracks which artifacts are complete:
| Flag | Meaning |
|---|---|
prd_complete |
A PRD exists and is approved |
architecture_complete |
Architecture doc and ADRs are done |
epics_complete |
Epics and stories are broken down |
implementation_signed_off |
Implementation is complete and peer-reviewed |
qa_signed_off |
QA sign-off is recorded — all P1/P2 bugs resolved |
release_complete |
Deployment is confirmed and release notes published |
legacy_inventory_complete |
Legacy codebase has been inventoried (modernization) |
knowledge_graph_complete |
Business rules and process graph built (modernization) |
The /sml-help skill reads these flags to give you context-aware guidance on what to do next.
Artifact output is written to sub-folders under output_path (configured in _superml/config.yml):
| Phase | Sub-folder | Config key |
|---|---|---|
| Product/BA, Architect, Team Lead | {output_path}/planning/ |
planning_artifacts |
| Developer | {output_path}/implementation/ |
implementation_artifacts |
| QA | {output_path}/qa/ |
qa_artifacts |
| Release | {output_path}/release/ |
release_artifacts |
| Modernization Lead | {output_path}/modernize/ |
modernize_artifacts |
All paths are created automatically by init. The output_path defaults to _superml-output/ but can be changed during setup.
Every significant event is appended to _superml/audit.log:
2025-01-15T09:00:00Z INIT project=acme-payments type=general
2025-01-15T09:01:00Z PERSONA user=alice persona=product
2025-01-16T14:22:00Z REENTER user=alice from=developer to=product
2025-01-16T14:30:00Z PERSONA user=alice persona=developer
The file is git-ignored and stays local to each team member.
npx @supermldev/smart-sdlc init- Creates
_superml/config.ymlwith project name, type, team names, and integration settings - Installs skills into
_superml/skills/ - Generates
.github/agents/and.github/skills/for GitHub Copilot users - Scaffolds
_superml/reference/folders for company documentation
npx @supermldev/smart-sdlc persona- Creates
_superml/persona.yml(git-ignored, stays local) - Captures: your name, role (
primary_persona), AI tool, skill level - If Copilot is selected and
.github/files don't exist yet, they are generated
| Type | Use When | First Skill |
|---|---|---|
general |
Feature work on an existing product | @sml-agent-analyst |
api |
Building or extending an API | @sml-agent-analyst → @sml-agent-architect |
greenfield |
Net-new product from scratch | @sml-agent-pm (Product Brief → PRD) |
modernization |
Migrating legacy systems | @sml-agent-scout → /sml-read-legacy-code |
npx @supermldev/smart-sdlc init # Set up project (run once)
npx @supermldev/smart-sdlc persona # Install or update your persona
npx @supermldev/smart-sdlc persona status # Show your current persona without wizard
npx @supermldev/smart-sdlc persona exit # Confirm and remove your persona
npx @supermldev/smart-sdlc reenter # Re-enter a prior phase with backward-entry guard
npx @supermldev/smart-sdlc help # Context-aware: what to do next
npx @supermldev/smart-sdlc list # Browse all skills
npx @supermldev/smart-sdlc meeting # Generate multi-persona meeting context
npx @supermldev/smart-sdlc update # Sync skills to latest version
npx @supermldev/smart-sdlc clean # Remove generated files# See who you are
npx @supermldev/smart-sdlc persona status
# Exit your current persona (e.g., before switching to another)
npx @supermldev/smart-sdlc persona exit
# Re-enter a prior phase (used for backward re-entry)
npx @supermldev/smart-sdlc reenter
# → choose phase → install persona for that phase → update artifact → exit → come backEach of these commands appends a log line to _superml/audit.log.