OnlyRead is a security control: a way to make an AI agent's browser session read-only. A bypass (a write that reaches a guarded site while enforce mode is on) is a security vulnerability.

Please report privately through GitHub: Security → Report a vulnerability on https://github.com/onlyreadai/onlyread. Do not open a public issue for a bypass.

Include the site, the action, how you observed that the write persisted (reload, another device, the other party), the time of the attempt, and whether OnlyRead showed its notice ("OnlyRead kept read-only"). Redact personal data.

Documented in the README under Limits and Roadmap: sites are guarded by host, not by account, Chrome's debugger banner, connections a site opened before guarding started, and agents that drive the screen from outside the browser, which are covered in Always mode only.