ChatGPT Personalization is a small static/browser + local-CLI project. It does not require API keys, account exports, a backend service, or runtime third-party Python packages. That reduces the attack surface, but it does not make published profiles or browser input automatically safe.

Security fixes are applied to the current main branch. Older commits, copied profiles, and third-party forks may not receive fixes.

Security-relevant reports include, for example:

• browser-builder behavior that could execute or inject untrusted content;
• unsafe import/export or file-handling behavior;
• a way to bypass the intended privacy boundary for profiles/local/;
• secret or sensitive-data exposure caused by repository tooling;
• CI or repository automation that could expose credentials or allow unintended writes.

A bad personalization result, inaccurate model answer, or ChatGPT product issue is not a security vulnerability in this repository.

Do not publish exploit details, credentials, or sensitive personal data in a public issue.

If GitHub shows a Report a vulnerability option for this repository, use that private reporting path. Otherwise, open a minimal issue stating that you need a private security contact, without including the exploit, secret, or sensitive payload.

GitHub private vulnerability reporting is a repository setting and may not always be enabled, so this policy does not assume that the button is present.

Do not include secrets or sensitive personal data in public presets, issues, pull requests, screenshots, or test fixtures. Personal JSON profiles belong in profiles/local/, which is ignored by Git by default.

The linter detects a small set of common credential-like patterns. It is a guardrail, not a security scanner, and cannot prove that a profile is safe to publish.

Please allow reasonable time to reproduce, assess, and fix a confirmed vulnerability before public disclosure. Security fixes should include a regression test when the issue can be reproduced safely.