AI_SCRUM/process/00_project_onboarding.md

Process: Project Onboarding (run once per new project)

Primary agent: .claude/agents/product-owner.md and .claude/agents/scrum-master.md

When to run

Before the first Backlog Refinement or Sprint Planning on a new project using this repo as a starting point. Re-run any step below if the project's stack or process changes materially.

Steps

1. Ask the user directly (do not guess) for:

   • What are we building, and who is the primary user? (feeds the product vision / first epics)
   • What is the tech stack (language, framework, database)?
   • What test framework(s) and how should tests be run?
   • Is code review required before merge? By whom (a person, or self-review by the agent)?
   • Is there a CI pipeline, or should one be set up? What must pass before merge?
   • What sprint length should we use (e.g., 1 week, 2 weeks)? This becomes fixed once chosen — do not change it sprint-to-sprint.
   • Any hard constraints (deadlines, compliance, performance, security requirements)?

2. Record the answers in project_config.md at the repo root (sprint length, stack, constraints).

3. Turn the review/test/CI answers into a concrete Definition of Done and overwrite the placeholder in templates/definition_of_done.md.

4. Create the initial product backlog structure in backlog/ (see backlog/backlog.md) from the product vision, seeding a first pass of epics under backlog/epics/ using backlog/epics/_epic_template.md.

5. Confirm the code/ folder is where all implementation work will live going forward.

Exit criteria

• project_config.md exists with sprint length and stack recorded.
• templates/definition_of_done.md no longer contains placeholder/generic content.
• At least one epic exists in backlog/epics/.

On completion: update state.md — phase → 1 (Backlog refinement), leading agent → product-owner, next action → refine the seeded epics toward “ready.” This is a natural pause point; confirm with the user before moving into backlog refinement.