Retrospective Analyzer

Analyze team retrospectives for insights

Instructions

1. Retrospective Setup

   • Identify sprint to analyze (default: most recent)
   • Check Linear MCP connection for sprint data
   • Define retrospective format preference
   • Set analysis time range

2. Sprint Data Collection

Quantitative Metrics

From Linear/Project Management:
- Planned vs completed story points
- Sprint velocity and capacity
- Cycle time and lead time
- Escaped defects count
- Unplanned work percentage

From Git/GitHub:
- Commit frequency and distribution
- PR merge time statistics  
- Code review turnaround
- Build success rate
- Deployment frequency

Qualitative Data Sources

1. PR review comments sentiment
2. Commit message patterns
3. Slack conversations (if available)
4. Previous retrospective action items
5. Support ticket trends

3. Automated Analysis

Sprint Performance Analysis

# Sprint [Name] Retrospective Analysis

## Sprint Overview
- Duration: [Start] to [End]
- Team Size: [Number] members
- Sprint Goal: [Description]
- Goal Achievement: [Yes/Partial/No]

## Key Metrics Summary

### Delivery Metrics
| Metric | Target | Actual | Variance |
|--------|--------|--------|----------|
| Velocity | [X] pts | [Y] pts | [+/-Z]% |
| Completion Rate | 90% | [X]% | [+/-Y]% |
| Defect Rate | <5% | [X]% | [+/-Y]% |
| Unplanned Work | <20% | [X]% | [+/-Y]% |

### Process Metrics
| Metric | This Sprint | Previous | Trend |
|--------|-------------|----------|-------|
| Avg PR Review Time | [X] hrs | [Y] hrs | [↑/↓] |
| Avg Cycle Time | [X] days | [Y] days | [↑/↓] |
| CI/CD Success Rate | [X]% | [Y]% | [↑/↓] |
| Team Happiness | [X]/5 | [Y]/5 | [↑/↓] |

Pattern Recognition

## Identified Patterns

### Positive Patterns 🟢
1. **Improved Code Review Speed**
   - Average review time decreased by 30%
   - Correlation with new review guidelines
   - Recommendation: Document and maintain process

2. **Consistent Daily Progress**
   - Even commit distribution throughout sprint
   - No last-minute rush
   - Indicates good sprint planning

### Concerning Patterns 🔴
1. **Monday Deploy Failures**
   - 60% of failed deployments on Mondays
   - Possible cause: Weekend changes not tested
   - Action: Implement Monday morning checks

2. **Increasing Scope Creep**
   - 35% unplanned work (up from 20%)
   - Source: Urgent customer requests
   - Action: Review sprint commitment process

4. Interactive Retrospective Facilitation

Pre-Retrospective Report

# Pre-Retrospective Insights

## Data-Driven Discussion Topics

### 1. What Went Well 
Based on the data, these areas showed improvement:
- ✅ Code review efficiency (+30%)
- ✅ Test coverage increase (+5%)
- ✅ Zero critical bugs in production
- ✅ All team members contributed evenly

**Suggested Discussion Questions:**
- What specific changes led to faster reviews?
- How can we maintain zero critical bugs?
- What made work distribution successful?

### 2. What Didn't Go Well
Data indicates challenges in these areas:
- ❌ Sprint velocity miss (-15%)
- ❌ High unplanned work (35%)
- ❌ 3 rollbacks required
- ❌ Team overtime increased

**Suggested Discussion Questions:**
- What caused the velocity miss?
- How can we better handle unplanned work?
- What led to the rollbacks?

### 3. Action Items from Data
Recommended improvements based on patterns:
1. Implement feature flags for safer deployments
2. Create unplanned work budget in sprint planning
3. Add integration tests for [problem area]
4. Schedule mid-sprint check-ins

Live Retrospective Support

During the retrospective, I can help with:

1. **Fact Checking**: 
   "Actually, our velocity was 45 points, not 50"

2. **Pattern Context**:
   "This is the 3rd sprint with Monday deploy issues"

3. **Historical Comparison**:
   "Last time we had similar issues, we tried X"

4. **Action Item Tracking**:
   "From last retro, we completed 4/6 action items"

5. Retrospective Output Formats

Standard Retrospective Summary

# Sprint [X] Retrospective Summary

## Participants
[List of attendees]

## What Went Well
- [Categorized list with vote counts]
- Supporting data: [Metrics]

## What Didn't Go Well  
- [Categorized list with vote counts]
- Root cause analysis: [Details]

## Action Items
| Action | Owner | Due Date | Success Criteria |
|--------|-------|----------|------------------|
| [Action 1] | [Name] | [Date] | [Measurable outcome] |
| [Action 2] | [Name] | [Date] | [Measurable outcome] |

## Experiments for Next Sprint
1. [Experiment description]
   - Hypothesis: [What we expect]
   - Measurement: [How we'll know]
   - Review date: [When to assess]

## Team Health Pulse
- Energy Level: [Rating]/5
- Clarity: [Rating]/5
- Confidence: [Rating]/5
- Key Quote: "[Notable team sentiment]"

Trend Analysis Report

# Retrospective Trends Analysis

## Recurring Themes (Last 5 Sprints)

### Persistent Challenges
1. **Deployment Issues** (4/5 sprints)
   - Root cause still unresolved
   - Recommended escalation

2. **Estimation Accuracy** (5/5 sprints)
   - Consistent 20% overrun
   - Needs systematic approach

### Improving Areas
1. **Communication** (Improving for 3 sprints)
2. **Code Quality** (Steady improvement)

### Success Patterns
1. **Pair Programming** (Mentioned positively 5/5)
2. **Daily Standups** (Effective format found)

6. Action Item Generation

Smart Action Items

Based on retrospective discussion, here are SMART action items:

1. **Reduce Deploy Failures**
   - Specific: Implement smoke tests for Monday deploys
   - Measurable: <5% failure rate
   - Assignable: DevOps team
   - Relevant: Addresses 60% of failures
   - Time-bound: By next sprint

2. **Improve Estimation**
   - Specific: Use planning poker for all stories
   - Measurable: <20% variance from estimates
   - Assignable: Scrum Master facilitates
   - Relevant: Addresses velocity misses
   - Time-bound: Start next sprint planning

Error Handling

No Linear Data

"Linear MCP not connected. Using git data only.

Missing insights:
- Story point analysis
- Task-level metrics
- Team capacity data

Would you like to:
1. Proceed with git data only
2. Manually input sprint metrics
3. Connect Linear and retry"

Incomplete Sprint

"Sprint appears to be in progress. 

Current analysis based on:
- [X] days of [Y] total
- [Z]% work completed

Recommendation: Run full analysis after sprint ends
Proceed with partial analysis? [Y/N]"

Advanced Features

Sentiment Analysis

# Analyze PR comments and commit messages
sentiment_indicators = {
    'positive': ['fixed', 'improved', 'resolved', 'great'],
    'negative': ['bug', 'issue', 'broken', 'failed', 'frustrated'],
    'neutral': ['updated', 'changed', 'modified']
}

# Generate sentiment report
"Team Sentiment Analysis:
- Positive indicators: 65%
- Negative indicators: 25%  
- Neutral: 10%

Trend: Improving from last sprint (was 55% positive)"

Predictive Insights

"Based on current patterns:

⚠️ Risk Predictions:
- 70% chance of velocity miss if unplanned work continues
- Deploy failures likely to increase without intervention

💡 Opportunity Predictions:
- 15% velocity gain possible with proposed process changes
- Team happiness likely to improve with workload balancing"

Experiment Tracking

"Previous Experiments Results:

1. 'No Meeting Fridays' (Sprint 12-14)
   - Result: 20% productivity increase
   - Recommendation: Make permanent

2. 'Pair Programming for Complex Tasks' (Sprint 15)
   - Result: 50% fewer defects
   - Recommendation: Continue with guidelines"

Integration Options

1. Linear: Create action items as tasks
2. Slack: Post summary to team channel
3. Confluence: Export formatted retrospective page
4. GitHub: Create issues for technical debt items
5. Calendar: Schedule action item check-ins

Best Practices

1. Data Before Discussion: Review metrics first
2. Focus on Patterns: Look for recurring themes
3. Action-Oriented: Every insight needs action
4. Time-boxed: Keep retrospective focused
5. Follow-up: Track action item completion
6. Celebrate Wins: Acknowledge improvements
7. Safe Space: Encourage honest feedback

References

• @$AIWG_ROOT/agentic/code/addons/aiwg-utils/rules/research-before-decision.md — Collect sprint data and previous retrospective action items before pattern analysis
• @$AIWG_ROOT/agentic/code/addons/aiwg-utils/rules/vague-discretion.md — Action items must be SMART (specific, measurable); avoid "improve process" without concrete steps
• @$AIWG_ROOT/agentic/code/addons/aiwg-utils/rules/human-authorization.md — Identify patterns and options; await team discussion before assigning retrospective action owners
• @$AIWG_ROOT/agentic/code/frameworks/sdlc-complete/skills/project-health-check/SKILL.md — Health metrics feed into retrospective data collection as quantitative evidence