Claude Code Session Statistics

Problem

As Claude Code sessions accumulate, developers usually have to review raw context one by one, making it hard to tell which sessions are worth revisiting and which tool-call or decision patterns keep recurring. CC Session Statistics organizes Claude Code session records into a selectable list, basic stats, and detailed reports, reducing manual scanning and fragmented notes.

How It Works

The skill parses $ARGUMENTS and enters different modes:
- no arguments or -l [limit]: lists available sessions, default 20
- -s <session>: generates basic stats for a specific session
- -d <session>: generates a detailed analysis for a specific session
- -a [limit]: aggregates recent sessions, default 10
- -h [limit]: shows history, default 20

The core step is to run analyze-session.mjs first, then generate the report from the returned data. For /vcs -d, the report must preserve the full structure; sections cannot be omitted just because a field is 0, empty, or has no result. Key findings and improvement suggestions must each include at least 3 items, and the complete report must be saved to a file with the Bash tool.

Boundaries

• The skill focuses on viewing, measuring, and reviewing Claude Code sessions; it does not replace project code review.
• Detailed analysis depends on script output; missing data should still be shown as 0 or 无 in the report.
• -d is better for deep dives into a single session, while -a or -h is better for cross-session trends.
• File saving depends on the Bash tool, so actual output availability depends on runtime permissions.