session-report

To generate an HTML report from Claude Code session transcripts, use the session-report skill to parse JSONL transcript data, deduplicate entries, and embed interactive JSON into a self-contained HTML dashboard summarizing token usage and cache behavior.

You can analyze Claude Code token usage and cache efficiency by processing session transcripts to summarize cache misses, token consumption, and subagent activity, which surfaces bottlenecks and unusually costly prompts.

The session report tool works with both recent and all-time usage analysis, allowing you to drill down into specific project sessions to detect anomalies and disproportionate spend across your entire Claude Code history.

The best way to identify expensive prompts and heavy subagent activity is to aggregate session transcript data into an interactive dashboard, highlighting cache breaks and costly subagent types to surface waste quickly.

Yes, you can audit subagent activity and token consumption without external dependencies because the skill operates independently by reading bundled HTML templates and embedding full report JSON directly into the output file.

Duplicate transcript entries appear in raw JSONL data, but the report generation process applies deduplication logic to preserve accurate token counts and cache efficiency metrics without inflating usage statistics.