Remix.run Logo
cedws 3 hours ago

While GitHub obsess over shoving AI into everything, the rest of the platform is genuinely crumbling and its security flaws are being abused to cause massive damage. Last week Aqua Security was breached and a few repositories it owns were infected. The threat actors abused widespread use of mutable references in GitHub Actions, which the community has been screaming about for years, to infect potentially thousands of CI runs. They also abused an issue GitHub has acknowledged but refused to fix that allows smuggling malicious Action references into workflows that look harmless.

GHA can’t even be called Swiss cheese anymore, it’s so much worse than that. Major overhauls are needed. The best we’ve got is Immutable Releases which are opt in on a per-repository basis.

imglorp an hour ago | parent | next [-]

Public service announcement

You can pin actions versions to their hash. Some might say this is a best practice for now. It looks like this, where the comment says where the hash is supposed to point.

      Old -->   uses: actions/checkout@v4
      New -->   uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4
There is a tool to sweep through your repo and automate this: https://github.com/mheap/pin-github-action
lijok an hour ago | parent | next [-]

The problem is actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 probably doesn’t do this same pinning, and the actions ecosystem is such an intertwined mess that any single compromised action can propagate to the rest

imglorp 41 minutes ago | parent | next [-]

Yes, true, but at least the fire won't spread through this one point. Hopefully all of your upstreams can be persuaded to pin also.

derfniw 42 minutes ago | parent | prev [-]

Well, it is a git commit hash of the action repo that contains the transpiled/bundled javascript.

Like: https://github.com/actions/checkout/tree/11bd71901bbe5b1630c...

So I'm pretty sure that for the same commit hash, I'll be executing the same content.

hobofan 31 minutes ago | parent [-]

This is true specifically for actions/checkout, but composite actions can have other actions as dependencies, and unless the composite action pins the versions of its dependencies, it is vulnerable for this attack.

This article[0] gives a good overview of the challenges, and also has a link to a concrete attack where this was exploited.

[0]: https://nesbitt.io/2025/12/06/github-actions-package-manager...

rtaylorgarlock 23 minutes ago | parent | prev | next [-]

My preferred tool to solve these issues is called 'gitlab'

righthand 9 minutes ago | parent [-]

CircleCI

TravisCI

Jenkins

scripts dir

Etc

kanzure 40 minutes ago | parent | prev | next [-]

I've always been worried about their backend changing and somehow named tags with a previous commit hash working for an attacker to give something you didn't expect for the commit hash.

woodruffw 39 minutes ago | parent | prev [-]

See also pinact[1], gha-update[2], and zizmor's unpinned-uses[3].

The main desiderata with these kinds of action pinning tools is that they (1) leave a tag comment, (2) leave that comment in a format that Dependabot and/or Renovate understands for bumping purposes, and (3) actually put the full tag in the comment, rather than the cutesy short tag that GitHub encourages people to make mutable (v4.x.y instead of v4).

[1]: https://github.com/suzuki-shunsuke/pinact

[2]: https://github.com/davidism/gha-update

[3]: https://docs.zizmor.sh/audits/#unpinned-uses

ljm 2 hours ago | parent | prev | next [-]

I worry that CI just got overcomplicated by default when providers started rocking up with templated YAML and various abstractions over it to add dynamic behaviour, dependencies, and so on.

Perhaps mixing the CI with the CD made that worse because usually deployment and delivery has complexities of its own. Back in the day you'd probably use Jenkins for the delivery piece, and the E2E nightlies, and use something more lightweight for running your tests and linters.

For that part I feel like all you need, really, is to be able to run a suite of well structured shell scripts. Maybe if you're in git you follow its hooks convention to execute scripts in a directory named after the repo event or something. Forget about creating reusable 'actions' which depend on running untrusted code.

Provide some baked in utilities to help with reporting status, caching, saving junit files and what have you.

The only thing that remains is setting up a base image with all your tooling in it. Docker does that, and is probably the only bit where you'd have to accept relying on untrusted third parties, unless you can scan them and store your own cached version of it.

I make it sound simpler than it is but for some reason we accepted distributed YAML-based balls of mud for the system that is critical to deploying our code, that has unsupervised access to almost everything. And people are now hooking AI agents into it.

Hasnep 2 hours ago | parent [-]

I'm trying out SelfCI [1] for one of my projects and it's similar to what you were describing. My whole CI pipeline is just a shell script that runs the actual build and test commands, I can write a script in another language like python if I need more complexity and I can run it all locally at any time to debug.

[1] https://app.radicle.xyz/nodes/radicle.dpc.pw/rad%3Az2tDzYbAX...

tasuki an hour ago | parent | prev | next [-]

> GHA can’t even be called Swiss cheese anymore, it’s so much worse than that.

That's a high bar though. Few things are better than Swiss cheese.

a-french-anon 2 hours ago | parent | prev [-]

If you want more ammo for your ranting (no offense meant, I also rant): an issue as massive as https://github.com/orgs/community/discussions/142308 lingering for years should do the trick.

llimllib an hour ago | parent [-]

It really feels like Firefox is not a supported browser on GitHub, I hit this and also find that much of the time the commit message is not correctly pulled from the PR description when that setting is enabled