Security teams should monitor for workflows that combine pull_request_target with cross fork pull requests, because that pattern can run untrusted code with elevated permissions. The highest risk appears when the workflow also checks out the PR head or runs on a self hosted runner. Webhooks and audit logs give the best chance to detect the abuse early.
What Makes This GitHub Actions Pattern Dangerous
The detection problem is less about any single workflow keyword and more about a dangerous combination of trust boundaries. In public repositories, pull request workflows can be used to trigger code that originates outside the repository owner’s trust boundary, so defenders should look for paths where untrusted content can influence a privileged run, especially when the workflow has write access or secrets access.
That combination becomes more serious when the workflow also checks out the contributor’s head branch or runs on infrastructure the repository owner controls. A workflow that can both observe attacker-controlled changes and then execute them with elevated context is the kind of pattern security teams should treat as suspicious by default.
For broader workflow hardening context, the GitHub and GitHub Actions Security Guide covers the surrounding controls that make these detections meaningful, including action pinning, runner hardening and workflow permission design.
What To Look For In Public Repositories
Security teams should prioritise detections that combine event type, repository visibility and execution context. In practice, the most useful signals include pull_request_target in public repositories, cross fork pull requests, explicit checkout of the PR head, self hosted runners, and workflows that request broad token permissions or access to sensitive secrets.
It also helps to watch for workflow logic that conditionally changes behavior based on file paths, labels or comments. Those patterns are not automatically malicious, but they are often used to smuggle attacker controlled logic into a trusted execution path. If the workflow can change what it runs based on input from the pull request, the trust boundary deserves close review.
The CI/CD Pipeline Identity Security Guide is a useful companion for understanding how tokens, runner trust and untrusted builds combine into a larger detection problem.
Which Telemetry Catches Abuse Early
Webhook activity and audit logs are the best early warning sources because they expose the event sequence around workflow execution, not just the final outcome. Teams should correlate pull request events, workflow dispatches, runner assignments, token use and secret access so they can reconstruct whether an untrusted contribution was able to influence a privileged job.
High-value detections also come from runner-side and repository-side mismatches. For example, a workflow that should only run on hosted runners but suddenly executes on self hosted infrastructure, or one that unexpectedly reaches secret-bearing steps after a forked contribution, should be treated as a potential abuse path until proven otherwise.
That is why a practical detection program benefits from supply-chain incident lessons such as GhostAction campaign 2025 and tj-actions/changed-files compromise 2025, because both show how workflow trust can be converted into secret exposure.
Risk and Threat Considerations
Public repository workflows are attractive because they sit at the intersection of contributor input, automation and privileged access. If a trusted workflow can be influenced by attacker-controlled code, the failure is not just code execution, it is the possibility of secret exposure, token abuse and downstream repository or infrastructure compromise.
Failure mechanism: A workflow accepts untrusted pull request content, then evaluates or checks out that content in a context that still has elevated permissions, secret access or runner trust.
Impact: An attacker can exfiltrate credentials, manipulate releases, tamper with build output or pivot into other systems that trust the workflow identity.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
MITRE ATT&CK addresses the attack and risk surface, while CIS Controls v8 and NIST SP 800-53 Rev 5 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| MITRE ATT&CK | T1190 — Exploit Public-Facing Application | Workflow abuse in public repos mirrors exposed trust-boundary exploitation. |
| Recommendation — Correlate exposed workflow entry points with abuse of trusted execution paths. | ||
| CIS Controls v8 | CIS-16 — Application Software Security | GitHub Actions workflows need secure review, hardening and abuse detection. |
| Recommendation — Review workflow logic, permissions and runner exposure for insecure automation paths. | ||
| NIST SP 800-53 Rev 5 | AU-2 — Event Logging | Detecting dangerous workflow use depends on capturing audit and webhook events. |
| AC-6 — Least Privilege | The core risk is elevated workflow permissions applied to untrusted PR content. | |
| SI-4 — System Monitoring | Monitoring suspicious workflow behavior requires alerting on anomalous execution patterns. | |
| Recommendation — Log workflow events, runner activity and privileged actions for investigation. Limit workflow permissions and secret access to the minimum required context. Monitor for suspicious workflow triggers, runner changes and secret access attempts. | ||
Practitioner Guidance
What to prioritise: Build detections around the combination, not the individual keyword. A plain pull_request_target alert is too noisy; the useful rule is pull_request_target plus fork origin plus checkout of attacker-controlled code plus access to secrets or a self hosted runner.
What to verify: Confirm whether the workflow can ever read repository secrets, create write-side effects, or invoke sensitive deployment steps after processing untrusted input. If it can, treat the repository as higher risk even when no incident has been confirmed.
Practitioner takeaway: The best detection strategy is to model where untrusted code crosses into trusted execution, then alert on workflows that turn that crossing into privileged action.
Related resources from NHI Mgmt Group
Deepen Your Knowledge
Free weekly newsletter
Subscribe to the NHI & AI Identity Journal
The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.
Bonus 33% off our NHI Course when you subscribe.
Reviewed and updated by the NHIMG editorial team on October 30, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org