What problem does it solve? Claude Code usage data is scattered across thousands of JSONL transcript files in ~/.claude/projects, making it hard to see where tokens, cache misses, and subagent costs actually go. This Skill consolidates those transcripts into a single explorable HTML report with findings and optimization suggestions. ## Core Features & Use Cases - Transcript Analysis: Runs a bundled Node.js analyzer that deduplicates API calls, attributes tokens to projects, skills, subagent types, and individual prompts, and detects cache breaks over 100k uncached tokens. - Self-Contained HTML Report: Embeds the JSON analysis into a bundled template with sortable tables, block-char project bars, a per-day session timeline, and expandable prompt drill-downs. - Findings & Recommendations: Fills the report with 3-5 data-backed takeaways (waste, anomalies, healthy signals) and concrete optimization callouts tied to specific rows. - Use Case: After a heavy week of Claude Code usage, run the report to discover that one project consumed 41% of tokens across 3 sessions and that cache-hit rate dropped below 85%, then adjust your workflow accordingly. ## Quick Start Ask the AI to generate a session report of my Claude Code usage for the last 7 days and save it to the current directory.
