The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script usage_summary.py to process session logs. This is standard behavior for a diagnostic tool and does not involve elevated privileges or dangerous commands.- [DATA_EXPOSURE]: The skill reads .jsonl session files located in ~/.claude/projects to aggregate token usage data. While these files contain session history, the script only extracts non-sensitive metadata (timestamps, model names, and token counts) for reporting purposes and does not exfiltrate the data.

claude-usage