Readit News logoReadit News
skywhopper · 5 months ago
I’m curious exactly what happened here. The 404media article isn’t detailed enough to be sure. My guess is the PR took advantage of some code injection possibilities in the GitHub Actions on the repo to grant the attacker admin access. But that’s a wild guess.
gruez · 5 months ago
>My guess is the PR took advantage of some code injection possibilities in the GitHub Actions on the repo to grant the attacker admin access. But that’s a wild guess.

Someone below mentioned the offending commit[1], which seems to be a doppelganger of another commit[2]. Maybe the exact commit message broke the automation?

[1] https://github.com/aws/aws-toolkit-vscode/commit/678851bbe97...

[2] https://github.com/aws/aws-toolkit-vscode/commit/d1959b99684...

QuinnyPig · 5 months ago
Exactly my position. I can’t realistically assess the potential scope of damage without a proper disclosure from AWS’s normally-excellent security team.
shdjhdfh · 5 months ago
Your article breathlessly blames AWS for being reckless while having no real facts about the compromise. The whole thing reads like click bait.
shdjhdfh · 5 months ago
The prompt 404 quotes in the article doesn't appear to exist anywhere in the git history for the repo they point to. It seems unlikely that Amazon would rewrite git history to hide this. Maybe the change was in a repo pulled in as a dependency.
shdjhdfh · 5 months ago
Ah, I think it might have been this, which was reverted and seems to have been pushed directly to master: https://github.com/aws/aws-toolkit-vscode/commit/678851bbe97...
Technetium · 5 months ago
I found a postmortem which seems to be well written: https://www.mbgsec.com/posts/2025-07-24-constructing-a-timel...
Kiboneu · 5 months ago
Copy-on-write filesystems should be the norm.

Another article came out earlier about dataloss from some vibecoding project and an automated snapshot setup would have mitigated this very issue.

blibble · 5 months ago
I guess they put their AI in charge of code review?