Agile Project Management Agent Rules
Project Context
You are managing or working within an agile software development team using Scrum or Kanban. Practices must support team velocity, predictability, and continuous improvement.
Issue Writing
- Write issue titles as user-facing outcomes, not technical tasks: `User can reset password via email` not `Implement password reset endpoint`.
- Include acceptance criteria as a checklist: `- [ ] User receives email within 60 seconds`, `- [ ] Link expires after 24 hours`.
- Add reproduction steps for bugs: environment, steps to reproduce, expected vs actual behavior, and any error messages.
- Label every issue by type: `bug`, `feature`, `tech-debt`, `spike`, `docs`, `security`.
- Set priority before placing an issue in the backlog — unrefined and unestimated issues belong in triage, not the sprint.
- Keep issues small enough to complete in 1–3 days — larger work must be decomposed before scheduling.
Task Decomposition
- Decompose vertically (thin end-to-end slices) not horizontally (all backend, then all frontend).
- Use the INVEST criteria: Independent, Negotiable, Valuable, Estimable, Small, Testable.
- Time-box research and spike tasks: a spike with no time limit becomes a feature.
- Create follow-up issues for known technical debt discovered during implementation — do not expand scope.
- Identify blocking dependencies during decomposition — anything that depends on external teams needs an earlier start.
Sprint Workflow
- Sprint duration: 1–2 weeks — choose one and stay consistent; changing cadence mid-project disrupts rhythm.
- Sprint planning: calculate team capacity (days × hours − PTO − ceremonies), then commit to a goal with supporting stories.
- Daily standup: what is blocked, what is next — under 15 minutes, update the board before the call.
- Sprint review: demo working software to stakeholders — only functionality that meets the definition of done counts.
- Sprint retrospective: capture one or two actionable improvements per sprint with assigned owners and due dates.
- Never auto-extend sprints — move incomplete work to backlog or next sprint explicitly after reviewing the cause.
Estimation
- Use story points on the Fibonacci scale (1, 2, 3, 5, 8, 13) — points measure relative complexity, not hours.
- Break any item over 8 points into smaller stories — large estimates signal unclear requirements.
- Use planning poker to prevent anchoring bias when estimating as a team.
- Establish reference stories as calibration anchors: "This 3-pointer is a standard CRUD endpoint with tests."
- Include time for code review, testing, documentation, and deployment in estimates — not just implementation.
- Track velocity over 3–5 sprints before using it for planning; do not compare velocity across teams.
Backlog Management
- Groom the backlog weekly: refine the top 20 items to be sprint-ready with acceptance criteria and estimates.
- Close or archive issues untouched for 90 days — if they were important, they would have been prioritized.
- Separate "must have" from "nice to have" — honest prioritization prevents backlog sprawl.
- Use milestones or epics to group issues by release or theme for roadmap visibility.
- Order the backlog by business value and technical risk — high-value, high-risk items first.
Definition of Done
- Code written and following project conventions.
- Linting and type checking pass.
- Unit tests written and passing.
- Code review approved by at least one team member.
- CI/CD pipeline passes all stages.
- Feature deployed to staging and smoke-tested.
- Acceptance criteria verified by the author or QA.
- Documentation updated if user-facing behavior changed.
- Issue transitioned to done and linked to the merged PR.
Communication & Transparency
- Write decisions in issues or ADRs, not in Slack threads that disappear.
- Tag relevant team members on blocking decisions rather than waiting for the next standup.
- Update issue status promptly — a stale board is worse than no board.
- Post weekly summaries to keep stakeholders informed without requiring attendance at team ceremonies.
- Escalate blockers on the same day they are identified — do not sit on impediments.
Metrics & Improvement
- Track cycle time (start to done) to surface bottlenecks in the workflow.
- Track escaped defects (bugs found after release) to improve testing practices.
- Review velocity trends over sprints — use it for planning, never as a performance metric.
- Run retrospective action items as first-class issues in the backlog with assigned owners.
- Measure lead time (idea to production) for strategic planning and commitment reliability.
Project Management
