GitHub will block the GitHub pull_request_target trigger by default in public repositories starting December 2, 2026, and any workflow that depends on it will fail until you explicitly allow it. If you maintain an open source project that runs automation on fork pull requests, you’ve got about two weeks to find those workflows and decide what to do with each one.
Here’s the short version, based on GitHub’s changelog:
- Deadline: permanent enforcement begins December 2, 2026.
- Who’s affected: public repositories on GitHub Free, Team and Enterprise Cloud.
- Who isn’t: private repositories, internal repositories and GitHub Enterprise Server.
- Already happened: test blackouts on November 9 and November 16, each from 16:00 to 18:00 UTC.
- Your move: allow the trigger through an Actions event policy, or switch the workflow to a safer trigger.
That list covers the what. The rest of this post covers the part that actually takes your time: working out which of your workflows break and how to fix them.
What GitHub pull_request_target does and why it keeps causing breaches
Workflows triggered by a pull request from a fork normally run with limited permissions and no access to your secrets. The pull_request_target trigger flips that. It runs in the context of your base repository, so it can read secrets and use a token with write access.
That’s handy for labeling bots and automated reviewers that need to comment on outside contributions. It turns dangerous when the workflow also checks out and runs code from the fork, or pipes attacker-controlled pull request text into a shell command. Researchers call these “pwn requests,” and according to Socket, they’ve been warned about since at least 2021. Socket ties variations of the pattern to supply chain incidents involving Nx, PostHog and TanStack.
GitHub already tightened actions/checkout in June to stop unsafe fork checkouts, Socket notes. The December change goes further because it blocks the trigger itself, including workflows that use it safely.
What changes on December 2, and what already happened
The rule rolled out in stages so maintainers could see the damage before it became permanent. First came evaluate mode, where workflows still ran but showed warnings. Then came two short blackouts, and now comes the real block.
| Date | What happens | Effect on your runs |
|---|---|---|
| Nov. 9, 16:00–18:00 UTC | First test blackout | Blocked runs failed with an error annotation |
| Nov. 16, 16:00–18:00 UTC | Second test blackout | Same failure, then back to warnings |
| Between blackouts | Evaluate mode | Runs execute, with warnings |
| Dec. 2, 2026 | Permanent block in affected public repos | Runs fail unless a policy allows them |
GitHub’s changelog doesn’t give an exact start time on December 2. Treat the whole day as live, and don’t leave your fixes for the final hour.
How to find out which of your repos are affected
Start by listing every workflow that uses the trigger. GitHub points to Insights for seeing which runs the rule would block, but Socket points out that this evaluate-mode view is reserved for Enterprise Cloud. Smaller open source projects will mostly need to check by hand.
A quick manual audit looks like this:
- Search each public repo’s
.github/workflowsfolder forpull_request_target. - Open every match and check whether it checks out or runs the pull request’s code.
- Look for PR titles, branch names or comment text passed straight into
run:steps. - Skim recent workflow runs to see which of these actually fire.
Steps two and three are where the real risk hides, so don’t skip them just because a workflow has “always worked.”
Allow it or replace it: which fix fits your workflow
You’ve got three realistic options. The right one depends on whether the workflow truly needs secrets or write access when an outside contributor opens a pull request.
| Option | Best for | Trade-off |
|---|---|---|
Switch to pull_request |
Build and test jobs that only need the fork’s code | Fork runs get no secrets, so label or comment steps may fail |
| Allow it with an Actions event policy | Triage bots and labelers that need base-repo permissions | You own the risk, so scope it tightly |
| Leave it blocked | Workflows nobody uses anymore | Anything still depending on it stops working |
If you do allow the trigger, GitHub says you can target specific workflow files rather than a whole repository. Socket adds that existing policies aren’t replaced. GitHub’s docs on controlling workflow execution cover the setup, though the changelog doesn’t spell out the exact menu path.
Locking down what automation can reach is a theme that keeps coming up. Our write-ups on the Optimizely supply chain attack and the Railway breach and the keys to rotate show how fast one over-trusted pipeline or leaked token becomes a bigger mess.
Frequently Asked Questions
Does the pull_request_target block affect private repositories?
No. GitHub says private repositories, internal repositories and GitHub Enterprise Server aren’t in scope.
Will my workflows break on December 2?
If they use the trigger in a public repo and no Actions event policy allows them, yes. Runs will fail with an error annotation, the same way they did during the two test blackouts.
Is pull_request_target being removed from GitHub Actions?
No. The trigger still exists. It’s blocked by default in affected public repos unless a policy explicitly allows it.
Can I just switch my workflow to pull_request instead?
Often, yes, for plain build-and-test jobs. It won’t work if the job needs secrets or write permissions on fork pull requests, because those runs don’t get them.
The bottom line before December 2
Don’t blanket-allow the trigger just to turn red builds green. Spend an hour this week auditing, move simple CI to pull_request, and keep scoped allow rules only for the handful of bots that truly need base-repo access. The block is a nuisance, but it targets an attack pattern that’s already been tied to real supply chain incidents.


Leave a Reply