caveman-stats

You can track real-time token usage by parsing local JSONL session logs. This skill reads directly from the active Claude Code session log to provide exact input and output token numbers for accurate performance monitoring.

Calculate token savings by comparing your current session's log data against a standard baseline. This skill computes the difference automatically, displaying the exact token savings achieved during your AI-assisted development interactions.

Yes, you need direct read access to the active local JSONL session log file. The skill parses these local logs deterministically to generate accurate usage reports without relying on external estimation methods.

Yes, you can audit tokens precisely by reading local JSONL logs. This eliminates the uncertainty of AI token estimation by providing exact, log-based reporting on actual input and output usage during your coding sessions.

AI token estimation often differs from actual consumption because it lacks log-level precision. By parsing local JSONL session logs directly, you get deterministic usage reports that reflect true input and output token metrics.

Trigger the token usage metrics by typing the command /caveman-stats into your chat interface. This parses the active session log and immediately generates a deterministic report of your current token consumption and savings.