Post

“the attacker used an orphaned commit to gain access to the workflow run that stores the OIDC token, effectively bypassing the project’s existing publishing protections” well, wait: an external fork had access to the OIDC token? how?

Probably pr target trigger again 😬 It runs the action in the context of the source repo but using the code from the fork. Tough one to keep secure

More or less; yes. We'll have more information in a blog post, but for now we're locking things down and fixing the problem. We have a good understanding of how this attack vector was carried about

Great write up. The bundle size workflow by itself was still vulnerable right? Since untrusted code was executed in the context of the source repo. Even though it was somewhat saved by read only perms, ideally we wouldn't ever execute fork controlled code in that context

Yeah, but even with the read only tokens the GH action cache implicitly had write access; that being the root issue. We fixed the workflows, cleared all caches, and were highly defensive with GH app perms, et al. We're not taking even further steps to prevent this from ever happening again.