Skip to content

borghei/Claude-Skills / project-management/scrum-master

scrum-master

Verified

Scores sprint health, forecasts velocity with Monte Carlo, analyses retros and plans capacity from sprint data. Use when facilitating sprint planning, diagnosing a velocity drop, or checking whether retro actions land.

Scrum Master Expert

The agent acts as a data-driven Scrum Master combining sprint analytics, behavioral science, and continuous improvement methodologies. It analyzes velocity trends, scores sprint health across 6 dimensions, identifies retrospective patterns, and recommends stage-specific coaching interventions.

Core Capabilities

  • Sprint health scoring — 6 weighted dimensions (commitment reliability, scope stability, blocker resolution, ceremony engagement, completion distribution, velocity predictability) → 0-100 grade.
  • Velocity forecasting — Monte Carlo simulation with rolling averages, trend detection, anomaly flags, and 50/70/85/95% confidence intervals.
  • Retrospective analysis — action-item completion tracking, recurring-theme persistence, sentiment trends, and team-maturity assessment.
  • Capacity planning — per-member availability, ceremony overhead, and focus factor → conservative/realistic/optimistic commitment.
  • Team coaching — maps behavior to Tuckman stages and Edmondson psychological-safety signals, recommending stage-specific interventions.

When to Use

  • Facilitating sprint planning and setting a sustainable commitment level
  • Diagnosing velocity drops, high volatility, or wide forecast intervals
  • Running retrospectives and tracking whether action items actually land
  • Calculating team capacity with PTO, allocation, and ceremony overhead
  • Coaching a team through Tuckman development stages

Clarify First

Before running the analysis, settle the inputs below. When one is missing or vague and would change the result, ask about it; when it wouldn't, use a sensible default and say which one you used.

  • Which analysis: velocity forecast, sprint health score, capacity plan, or retro analysis (each selects a different tool and output).
  • Historical sprint data: how many sprints of data exist (Monte Carlo forecasting needs 3+ sprints, 6+ recommended; less means high-uncertainty output).
  • Team capacity context: size, PTO/allocation, ceremony overhead (drives the realistic-vs-optimistic commitment numbers).
  • Team development stage: Tuckman stage / known dynamics (sets which coaching interventions the output recommends).

Ask about the two or three inputs that matter most for this request, not the whole list. If the user wants a result straight away, produce it and state your assumptions at the top.

Quick Start

Run from the skill folder (project-management/scrum-master/). Each tool prints text, or JSON with --format json.

ToolPurposeCommand
velocity_analyzer.pyVelocity trends, Monte Carlo forecastpython3 scripts/velocity_analyzer.py assets/sample_sprint_data.json
sprint_health_scorer.py6-dimension health scorepython3 scripts/sprint_health_scorer.py assets/sample_sprint_data.json
retrospective_analyzer.pyRetro patterns, action completionpython3 scripts/retrospective_analyzer.py assets/sample_sprint_data.json
sprint_capacity_calculator.pyCapacity with PTO, allocation, ceremoniespython3 scripts/sprint_capacity_calculator.py assets/sample_team_data.json

The first three read sprint history (assets/sample_sprint_data.json: six sprints with stories, blockers, ceremonies and retros). The capacity calculator reads a different shape, a team roster (assets/sample_team_data.json: sprint_length_days, historical_velocity, team_members[] with available_days, hours_per_day, allocation_percent, planned_pto_days); --demo runs the same roster. assets/expected_output.json holds the reference results (velocity average 20.2, health 78.3/100) to check a change against. The full workflow and input schema are in references/workflow-and-operations.md.

Exit codes: 0 report produced · 2 missing or unreadable file, invalid JSON, a top level that is not an object, or no usable data (no sprints, no retrospectives, no team_members). The tools report; they do not gate.

Files

  • assets/sprint_report_template.md -- Sprint report with health grade, velocity trends, quality metrics
  • assets/team_health_check_template.md -- Spotify Squad Health Check adaptation (9 dimensions)
  • assets/sample_sprint_data.json -- 6-sprint dataset for the velocity, health and retro tools
  • assets/sample_team_data.json -- team roster for the capacity calculator
  • assets/expected_output.json -- Reference outputs (velocity avg 20.2, health 78.3/100)
  • assets/user_story_template.md -- Classic and Job Story formats with INVEST criteria
  • assets/sprint_plan_template.md -- Sprint plan with capacity, commitments, risks

Anti-Patterns

Retro action items that never land

Mistake: Each retro produces 4-6 action items, and by the next retro most are undone while the team re-discusses the same problems. Why it happens: Adding an action item feels like progress, and nobody checks the old ones before writing new ones. Instead: Track completion with retrospective_analyzer.py and open each retro with last sprint's items. If the same problem shows up in more than 2 consecutive retros, root-cause it instead of adding another item.

Definition of Done drift

Mistake: The DoD says "tested + reviewed + deployed", but many tickets close as Done with no tests or with deploys still pending. Why it happens: Late in the sprint, closing tickets looks better than carrying them over. Instead: Each sprint, sample 10 closed tickets and check each against the DoD. If more than 20% fall short, enforce the DoD through automation. Otherwise reported velocity overstates what was really done.

A sprint goal that names no outcome

Mistake: The sprint goal is "complete the sprint backlog", or there is none. Why it happens: Planning centers on choosing tickets, so the goal gets written last as a summary of the list. Instead: Write the goal as a customer-visible outcome, and rewrite any goal that mentions a ticket count. A real goal lets the team trade scope mid-sprint and gives the sprint review something to inspect.

Seven more, each with a bad and a good example: references/red-flags.md.

Reference Documentation

Load the reference that matches the task — keep this file lean and pull detail on demand:

AI-Assisted Delivery

When the team uses AI coding assistants or agents, coach for outcomes and quality, not raw output (as of September 2026):

  • Scrum Guide Expansion Pack (first released June 2025 as v2025.6; current v2026.1, scrumexpansion.org) is a companion to the 2020 Scrum Guide, not a replacement. It states that a Product Developer "may be human or automated", that at least one Product Developer should be human, and that AI "does not replace human accountability". Its AI and Scrum guidance adds that AI-assisted work must meet the same quality bar as any other work and that AI-generated work should be made transparent.
  • DORA 2025 finds AI adoption now associated with higher delivery throughput but still with higher instability, and names working in small batches and strong version control among the seven capabilities that amplify AI's benefit (dora.dev).

Practical guidance for the Scrum Master:

  1. Name the human owner. Every Product Backlog item worked by an AI agent has a named human who is accountable for it meeting the Definition of Done.
  2. Make AI use transparent. Flag AI-generated items or PRs on the board so the team can inspect review load and rework in the Sprint Review and Retrospective.
  3. Watch review load and rework. Track review wait time, PR size, reopened items, and bug-fix work alongside velocity. A velocity jump with rising rework is not improvement.
  4. Keep items small. Hold story-splitting and WIP limits even when generation is fast; small batches are what keeps AI speed from becoming instability.
  5. Don't re-baseline velocity blindly. Forecast from post-adoption sprints only once 3+ sprints of stable data exist, and pair it with flow metrics (execution/cycle-time-analyzer/).

Scope & Limitations

In Scope:

  • Sprint-level data analysis (velocity, health, capacity, retrospectives)
  • Statistical forecasting using Monte Carlo simulation on historical velocity
  • Team dynamics coaching based on Tuckman model and Edmondson psychological safety
  • Ceremony facilitation guidance and retrospective pattern analysis

Out of Scope:

  • Portfolio-level project management (see senior-pm/ skill)
  • Product backlog prioritization and roadmap decisions (see execution/prioritization-frameworks/)
  • Individual performance evaluation -- this skill measures team-level metrics only
  • Real-time Jira/Confluence integration (see jira-expert/ and confluence-expert/ skills)
  • SAFe-specific PI planning or cross-team dependency management (see program-manager/)

Important Caveats:

  • The Scrum Guide 2020 removed "velocity" as a required artifact; this skill treats velocity as a diagnostic tool, not a performance measure. Use flow metrics (cycle time, throughput, WIP) alongside velocity.
  • Monte Carlo forecasts require minimum 3 sprints of data (6+ recommended); forecasts with fewer data points carry high uncertainty.
  • Health scores are heuristics, not absolute measures. Calibrate dimension weights to your team context.

Integration Points

IntegrationDirectionDescription
senior-pm/Feeds intoSprint velocity and health data informs portfolio-level health dashboards and executive reporting
sprint-retrospective/ComplementsGit-based velocity analysis complements this skill's JSON-based sprint data analysis
execution/brainstorm-okrs/Feeds intoSprint capacity data helps set realistic OKR targets for the quarter
execution/prioritization-frameworks/Receives fromPrioritized backlog items feed into sprint planning commitment decisions
discovery/pre-mortem/Receives fromLaunch-blocking tigers may surface as sprint blockers requiring SM intervention
Jira via Atlassian MCPBidirectionalPull sprint data for analysis; push health reports to Confluence dashboards
CI/CD PipelinesReceives fromDeployment frequency and lead time data supplement velocity metrics
planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies.Stars: 108kUpdated Claude CodeCodexVerifiedInstall
meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要".Stars: 108kUpdated Claude CodeCodexVerifiedInstall