What GitHub's pull_request_target changes break in the 1,000 most-starred repositories

GitHub is modifying its pull_request_target feature this year, impacting the 1,000 most-popular repositories on the platform. I developed a free online checker, prt-check, using my AI system (HAL) to analyze these repositories. To check your own repository, set up the prt-check action or run the provided Python script locally with no additional dependencies. The two key changes, effective from July 20, 2026, and November 2, 2026, respectively, are:

1. Starting July 20, 2026, actions/checkout no longer checks out fork pull requests in privileged workflows unless explicitly allowed via an Actions policy.

2. Starting November 2, 2026, GitHub will block pull_request_target on public repositories without an Actions policy permitting it.

Out of the 1,000 scanned repositories, 269 (26.9%) run at least one workflow on pull_request_target, while 9 (0.9%) utilize the new trigger in a privileged workflow, checking out fork code. These 9 repositories are the most urgent to address, as the guard refuse to check out their code. Additionally, 4 repositories manually opted-in to use the new guard (allow-unsafe-pr-checkout: true), but this is only safe if no subsequent steps execute the fork's files.

Nine repositories fetch pull request code via alternative methods that bypass the new guard. Eight repositories run AI or review actions on pull_request_target, which will stop functioning on November 2, 2026, unless the trigger is explicitly allowed. The most commonly used actions in these workflows are actions/github-script (125), actions/labeler (50), actions/create-github-app-token (25), and several others, with a total of 9,328 workflow files analyzed.