Custom Usage Insights

Run deep analytics across all Claude Code sessions (not just the last 6).

Usage

/my-insights                              # Full run (uses cache, fast on re-runs)
/my-insights --limit 20                   # Test with 20 sessions
/my-insights --project sandbagging        # Filter by project name
/my-insights --since 30                   # Last 30 days only
/my-insights --dry-run                    # Preview batch plan
/my-insights --report-only               # Regenerate report from cached facets
/my-insights --force                      # Ignore cache, reprocess everything

What It Does

1. Extracts clean transcripts from all session JSONL files
2. Analyzes each session via Gemini (goal categories, outcomes, friction, etc.)
3. Generates an HTML report with charts and per-project breakdowns

Facets are cached per-session — re-runs only process new/changed sessions.

Execution

python scripts/insights/run_insights.py $ARGUMENTS

Report output: ~/.claude/custom-insights/report.html