Analyzing Development Patterns

Bash
# Analyze last epic's patterns
git log --grep="MKPLS-XXX" --oneline --stat
/analyze-chat --epic=MKPLS-XXX

# Generate improvement recommendations
echo "## Process Analysis: MKPLS-XXX" > analysis.md
echo "Evidence: [findings]" >> analysis.md
echo "Recommendations: [actions]" >> analysis.md

Progress:

• Gather Data: Git history + conversation transcripts + metrics
• Identify Patterns: Success patterns vs anti-patterns
• Extract Insights: Root causes with evidence
• Generate Recommendations: Specific, actionable improvements
• Create Report: Structured findings with priorities
• Track Impact: Measure before/after effectiveness

1. Data Collection

Git History Analysis:

Bash
# Epic commits and patterns
git log --grep="MKPLS-XXX" --oneline --stat
git branch -a | grep MKPLS-XXX

# Quality trends
git log --format="%s" | grep "tests pass"

Conversation Analysis:

Bash
# Agent effectiveness and tool usage
/analyze-chat --date-range="2024-01-01,2024-01-15"

# Extract metrics:
# - Agent delegation patterns
# - Skill adoption rates  
# - Tool usage ratios
# - Cache efficiency

Metrics Collection:

• Test coverage trends
• Build success rates
• Time to resolution
• Rule adherence frequency

2. Pattern Recognition

Success Patterns (reinforce):

• Effective TDD workflows
• Clean agent delegation
• High rule adherence
• Strong cross-repo coordination

Anti-Patterns (address):

• Repeated rule violations
• Inefficient tool usage
• Communication breakdowns
• Quality regressions

3. Report Generation

Template structure:

Markdown
# Process Analysis: [Scope]

[Key findings and impact]

Pattern: [Name]

Evidence: [Data points] Frequency: [X instances] Impact: [Measurable outcome] Recommendation: [How to reinforce]

Pattern: [Name]

Evidence: [Data points] Root Cause: [Analysis] Impact: [Cost/delay] Recommendation: [Specific fix]

• HIGH: [High impact, low effort]
• MEDIUM: [Medium impact]
• LOW: [Nice to have]

Example 1: Agent Effectiveness Analysis Input: /analyze-chat --epic=MKPLS-211 Output:

Agent Usage:
- backend-specialist: 25/30 (83%) ⚠️ Dominant
- test-specialist: 3/30 (10%) ⚠️ Underused

Issue: Test agent underutilized
Evidence: 7 test files modified by backend-specialist
Root Cause: Unclear delegation rules
Action: Update agent-delegation.md with "test-only tasks → test-specialist"
Expected Impact: 35% test agent utilization, improved test quality

Example 2: TDD Adherence Pattern Input: Git history showing test-first violations Output:

Anti-Pattern: Tests written after implementation
Frequency: 8/12 features (67%)
Evidence: Implementation commits precede test commits
Impact: 3 bugs found in production, 2 days rework
Recommendation: Add pre-commit hook reminder "Run tests first"

Example 3: Tool Usage Optimization Input: High Bash:Read ratio (4:1) Output:

Anti-Pattern: Excessive file searching
Evidence: Bash:Read ratio 4:1 (expected <2:1)
Root Cause: Not reusing context efficiently
Action: Update rules for context reuse patterns
Expected Impact: 40% faster responses, lower token usage

Evidence-Based Insights:

• Quantify patterns with specific counts/percentages
• Link recommendations to measurable outcomes
• Show before/after comparisons when possible

Actionable Recommendations:

• Specify exact rule changes with line numbers
• Propose concrete workflow modifications
• Suggest new skills/agents for recurring needs

Balanced Perspective:

• Celebrate successes alongside improvements
• Consider team context and constraints
• Prioritize high-impact, low-effort changes

Incremental Improvement:

• Focus on 1-2 key changes per cycle
• Track adoption and effectiveness
• Iterate based on results

❌ Analysis paralysis: Don't spend more time analyzing than the work took ❌ Vague recommendations: "Improve testing" isn't actionable ❌ Change overload: Too many recommendations at once ❌ Problem-only focus: Balance issues with successes ❌ Ignoring context: Consider team capacity and priorities