pushback reads your coding-agent transcripts, which can hold client work, personal data and credentials.

• extract, audit, report and export run locally and send nothing.
• label and rules send message text to the model provider you configure, and ask before they do. rules --prompt-file lets you avoid that entirely.
• The export's secret scrub is best effort. Read an export before sharing it.

To report a vulnerability, for example a path where message text could leave your machine without that prompt, use GitHub's private vulnerability reporting on this repository (Security tab → Report a vulnerability). Please don't open a public issue for it.