name: session-analysis-patterns description: Claude Code session analysis, tool call patterns, efficiency metrics, and anti-pattern detection

Session Analysis Patterns

Session Metrics

Anti-Pattern Catalog

Efficiency Scorer (0-100)

def calculate_efficiency(session):
    score = 100

    # Penalties
    score -= session.repeated_reads * 5       # -5 per repeated read
    score -= session.blind_edits * 15         # -15 per blind edit
    score -= session.wide_greps * 3           # -3 per wide grep
    score -= session.idle_agents * 10         # -10 per idle agent
    score -= session.cascade_fails * 8        # -8 per cascade fail

    # Bonuses
    score += session.parallel_calls * 2       # +2 per parallel batch
    score += session.first_pass_success * 5   # +5 per first-try success

    return max(0, min(100, score))

Session Comparison

| Metrik | Session A | Session B | Delta |
|--------|-----------|-----------|-------|
| Total tokens | 45,000 | 120,000 | +167% |
| Tool calls | 12 | 38 | +217% |
| Time (min) | 3 | 12 | +300% |
| Tasks completed | 2 | 2 | 0% |
| Efficiency score | 88 | 42 | -52% |

Braintrust Log Analysis

# Session analizi
btca sessions list --last 7d
btca session <id> --metrics

# Pattern detection
btca session <id> --anti-patterns

# Cost analysis
btca sessions list --sort cost --desc

Checklist

• Session token usage tracked
• Anti-pattern detection aktif
• Efficiency score > 70
• Repeated reads < 2
• Blind edits = 0
• Parallel opportunities exploited
• Agent spawn ROI positive

Anti-Patterns

• Her session'ı analiz etmek (sadece outlier'ları analiz et)
• Metric'lere takılıp task quality'yi ihmal etmek
• Hız için correctness'tan ödün vermek