| name | scrum-master |
| description | Scrum facilitation: sprint ceremonies, retrospective formats, removing blockers, velocity coaching, scaling scrum across multiple teams — practical Scrum Master patterns |
Scrum Master Skill
When to activate
- Facilitating sprint ceremonies (planning, standup, review, retro)
- Running retrospectives that surface real issues (not just "what went well")
- Helping a team that's struggling with velocity, scope, or commitment
- Scaling from one team to multiple teams (Scrum of Scrums, SAFe basics)
- Coaching a new team on Scrum practices
- Writing agendas for sprint ceremonies
When NOT to use
- Jira configuration — use the jira-expert skill
- Product roadmap decisions — that's the PM's domain
- Engineering technical decisions — not a Scrum Master concern
- Replacing a real Scrum Master for teams in conflict — human facilitation needed
Instructions
Sprint retrospective
Design a retrospective for [team context].
Team size: [X people]
Sprint length: [2 weeks]
Team health: [healthy / some tension / struggling]
Last retro: [what format / what came out of it]
This sprint's standout event: [incident / delivery pressure / team change / none]
Choose a retrospective format based on context:
1. Start / Stop / Continue (default, all-team contexts):
- Start: what should we begin doing?
- Stop: what should we stop doing?
- Continue: what's working that we should protect?
Duration: 60 min for a 2-week sprint
2. 4Ls (after a tough sprint):
- Liked: what did you enjoy?
- Learned: what did you discover?
- Lacked: what was missing?
- Longed For: what do you wish had been different?
3. Sailboat (for teams feeling directionless):
- Wind (propelling us): what's helping us move forward?
- Anchors (slowing us): what's holding us back?
- Rocks (risks ahead): what could sink us?
- Sun (destination): what are we sailing toward?
4. Timeline (after incidents or major delivery):
- Map the sprint on a timeline
- Mark high points and low points as a team
- Discuss what caused each peak and valley
- Identify patterns
Facilitation guide for [chosen format]:
1. Set the stage (5 min): psychological safety framing
2. Gather data (15 min): silent stickies on the board
3. Insights (20 min): group stickies into themes, discuss
4. Decide what to do (15 min): vote on top 2-3 action items
5. Close (5 min): confirm owners and due dates for each action
Rule: no retro ends without a named owner and due date for each action item. "The team will..." = nobody will.
Generate the full agenda for my specific context.
Standup facilitation
Improve our daily standup.
Current standup: [describe — how long, format, problems]
Team size: [X people]
Remote / in-person / hybrid: [specify]
Common problems: [runs too long / people not present / no blockers shared / status report instead of sync]
Standup formats:
Classic 3 questions (per person):
1. What did I do yesterday?
2. What am I doing today?
3. Any blockers?
Problem: becomes a status report — people talk AT the Scrum Master, not to each other.
Walking the board (better for in-progress focus):
- Look at each "In Progress" item, not each person
- "Who's working on this? Any blockers?"
- Focuses on finishing, not starting
- Better for Kanban-adjacent teams
Two-question model (leaner):
1. What am I working on?
2. Do I need help?
No "yesterday" = cuts standup to < 10 minutes with < 10 people
Remote standup tips:
- Use a shared board (Jira, Linear) on screen — prevents abstract status reports
- Start on time, end on time — latecomers join without recap
- Blockers go in Slack async; standup is for coordination, not solving
Common standup anti-patterns to fix:
- "No blockers" every day → blockers exist; people aren't comfortable sharing
Fix: ask "what would make you go faster?" instead
- One person talks for 5+ minutes → use a timer (2 min/person)
- No one moves their tickets afterward → blockers or tickets are wrong
Redesign my standup for my team's context.
Velocity coaching
Help improve team velocity.
Current velocity: [X story points / sprint average last 3 sprints]
Sprint length: [2 weeks]
Team size: [X engineers]
Known issues: [scope creep / unrealistic estimates / interruptions / tech debt / unclear stories]
Velocity diagnosis framework:
Step 1 — Distinguish types of velocity problems:
a) Commitment problem: team commits to X, delivers Y < X → planning is broken
b) Estimation problem: team delivers X but stories keep being re-estimated higher mid-sprint
c) Interruption problem: unplanned work (bugs, incidents, slack requests) eating capacity
d) Delivery problem: stories sit "In Progress" for most of the sprint
Step 2 — Measure the real issue:
- Interruption rate: track unplanned work added mid-sprint for 3 sprints. If > 20% of committed work, that's the problem — not estimation.
- Cycle time: if stories take > 5 days on average, WIP limit is too high
- Commitment rate: committed / delivered over last 3 sprints
Step 3 — Interventions by problem type:
a) Commitment: run sprint planning with the team, not for them. Stop committing to unrefined stories.
b) Estimation: run a pointing calibration session (compare past estimates to actual)
c) Interruptions: budget for interruptions (reserve 20% of velocity for unplanned work)
d) Cycle time: enforce WIP limit of 2 stories per engineer maximum
Step 4 — Don't optimise velocity directly:
Velocity is a planning tool, not a performance metric. A team that does 40 points of meaningful work is better than one doing 60 points of low-value work.
Diagnose my team's velocity issue and recommend the single highest-leverage change.
Scrum ceremonies calendar
Design the sprint ceremonies calendar for [team].
Team: [X people — engineers, PM, designer, QA]
Sprint length: [2 weeks]
Time zone: [all same / distributed]
Working hours overlap: [X hours/day of overlap]
Two-week sprint ceremony schedule:
Monday, Sprint Start:
- 10:00 — Sprint Planning (2 hours)
Who: Full team
Goal: Sprint backlog agreed, sprint goal set
Daily Monday–Friday:
- 09:30 — Daily Standup (15 min max)
Who: Full team
Format: Walk the board
Friday of Week 1:
- 15:00 — Backlog Refinement (1 hour)
Who: PM + 2-3 engineers (rotating)
Goal: Top 10 backlog items estimated and clarified
Thursday of Week 2 (day before sprint ends):
- 14:00 — Sprint Review (1 hour)
Who: Full team + stakeholders
Goal: Demo completed work, gather feedback
- 15:30 — Retrospective (1 hour)
Who: Team only (no stakeholders)
Goal: 2-3 actionable improvements for next sprint
Friday of Week 2 (Sprint End):
- Team completes remaining work
- PM closes sprint in Jira, prepares velocity chart
Design the calendar for my team size, time zone, and working style.
Example
User: Our retrospectives always end with the same 3 action items ("write better tickets", "communicate more", "reduce scope creep") and nothing ever changes. How do I fix this?
Claude's diagnosis:
These 3 action items are process platitudes — they name symptoms without addressing causes, and they have no owner. That's why nothing changes. The retro is doing the right ceremony, but the wrong thing.
Root cause of "write better tickets":
This means: tickets that go into the sprint are underspecified, causing rework or wrong implementation.
Real question to ask: "Which ticket in the last 3 sprints caused the most rework? Why was it unclear? Who wrote it? What was missing?"
Specific action: "Before next sprint planning, PM reviews all tickets with the engineer who'll own them — takes 30 min every Thursday."
Root cause of "communicate more":
This is usually a symptom of unclear ownership, not a communication failure.
Real question: "In the last sprint, when did you not know what to do next and had to ask someone? Why weren't you told earlier?"
Specific action: "Add a 'dependencies' field to every story > 3 points — who else needs to be involved?"
Root cause of "reduce scope creep":
This means: items are added mid-sprint without removing others.
Real question: "Who approved the mid-sprint additions in the last 3 sprints? Was it urgent or avoidable?"
Specific action: "New item added mid-sprint requires explicit PM + tech lead sign-off AND removal of equivalent points."
Retro format change:
Stop using Start/Stop/Continue. Switch to a Timeline retro — go back through the last sprint day by day. This forces specific examples rather than generic complaints, and specific examples have specific causes.
