See [AGENTS.md](./AGENTS.md) for the full Scrum process this agent must follow in this repository, including the sub-agent orchestration order. That file is the canonical source of instructions.

**Read [`state.md`](./state.md) FIRST, before anything else, every session.** It says which phase the project is in, which sub-agent leads right now, and the next action — don't ask the user what to do next or infer it from chat history when `state.md` already says. If it doesn't exist, the project hasn't been touched: treat it as Phase 0 and start with onboarding below.

**Start every new project with `process/00_project_onboarding.md`** — it has the agent ask the user about stack, tests, review process, CI, and sprint length, then fills in `project_config.md` and `templates/definition_of_done.md`. Do not build on the generic placeholders.

Quick pointers:
- Current state (read first, update last): [`state.md`](./state.md)
- Sub-agents (Claude Code auto-discovers these): [`.claude/agents/product-owner.md`](./.claude/agents/product-owner.md), [`.claude/agents/scrum-master.md`](./.claude/agents/scrum-master.md), [`.claude/agents/dev-team.md`](./.claude/agents/dev-team.md), [`.claude/agents/qa-tech-debt.md`](./.claude/agents/qa-tech-debt.md), [`.claude/agents/release-planner.md`](./.claude/agents/release-planner.md)
- Step-by-step lifecycle: [`process/`](./process/)
- Fill-in templates: [`templates/`](./templates/)
- Live product backlog: [`backlog/backlog.md`](./backlog/backlog.md) + [`backlog/epics/`](./backlog/epics/) + current sprint at [`backlog/sprints/`](./backlog/sprints/)
- Impediment / tech-debt / process-improvement logs: [`logs/impediment_log.md`](./logs/impediment_log.md), [`logs/technical_debt_log.md`](./logs/technical_debt_log.md), [`logs/process_improvement_log.md`](./logs/process_improvement_log.md)
- Project settings: [`project_config.md`](./project_config.md)
- Codebase lives in: [`code/`](./code/)