TL;DR: Akeyless Security Research shows an automated campaign used 4,837 compromised GitHub accounts to probe 61,764 repositories and inject a 1,500-line shell payload into at least 1,806 workflows, harvesting static CI/CD secrets, cloud IAM credentials and AI developer tokens before deleting logs. Standing secrets and weak workflow governance turn a single token leak into a fast-moving cloud breach.
At a glance
What this is: NHI Mgmt Group's analysis, drawing on Akeyless's research: Automated GitHub abuse shows that static CI/CD secrets and weak workflow governance let attackers harvest credentials, pivot into cloud IAM and erase traces in minutes.
Why it matters: IAM, NHI and DevSecOps teams need to treat CI/CD runners as high-value identity brokers because one compromised workflow account can expose cloud, database and AI credentials at machine speed.
By the numbers:
- In at least 1,806 confirmed repositories, attackers injected a 1,500-line shell payload into GitHub Actions workflows before reverting commits and deleting run logs.
- The campaign leveraged 4,837 compromised GitHub accounts to open probe pull requests across 61,764 repositories.
- The first ~26 hours saw 61,839 probe pull requests across 61,764 repositories.
- At approximately 12:45 UTC on November 10, 2026, the campaign began with automated repository intrusions.
👉 Read Akeyless's analysis of the GitHub Actions secret-harvesting campaign
Context
Standing secrets in CI/CD create an identity problem as much as a tooling problem. A workflow runner that can read long-lived keys, cloud credentials and AI tokens becomes a credential concentration point, so a stolen developer account can reach far beyond source code.
The failure mode is not just initial compromise. Once an attacker can write to a repository or workflow file, the pipeline itself becomes the place where secrets are collected, privilege is expanded and evidence is erased. That makes secrets sprawl in pipelines a governance issue, not a local configuration issue.
Key questions
Q: What is the safest way to handle GitHub Actions secrets at runtime?
A: The safest runtime handling is to inject GitHub Actions secrets only into the specific job or step that needs them, keep them masked, and avoid passing them through logs, artefacts or shared environment variables. Rotate secrets regularly and enforce least privilege from a central control plane so scoped credentials can be revoked quickly if a runner is compromised.
Q: Why do CI/CD secrets create such a large blast radius in supply chain attacks?
A: CI/CD secrets are often shared across build, publish, and cloud tasks, so one exposed token can touch many systems at once. In this incident, the same compromise path reached GitHub, npm, AWS, GCP, Azure, Vault, SSH, and database credentials. That makes token scope and runner isolation central to blast-radius reduction.
Q: How does OIDC federation replace static cloud credentials in pipelines?
A: OIDC federation replaces stored cloud keys by letting the pipeline exchange a signed OIDC token for short-lived cloud access. The cloud trusts claims such as repository, branch, job and audience, so the workflow can assume a role without a long-lived secret. That removes exposed static credentials, narrows blast radius, and makes revocation easier through token policy and trust configuration.
Q: What should teams do first when workflow logs disappear after a CI compromise?
A: Teams should first preserve and isolate the CI system, then collect immutable evidence from the source repository, runner host, orchestration layer and central log platform. Missing run numbers usually mean the attacker altered or deleted job metadata, so compare commit history, pipeline definitions, artefact timestamps and authentication events to rebuild the execution chain before credentials are rotated or runners are rebuilt.
Technical breakdown
Why GitHub Actions runners become credential concentration points
A CI/CD runner often inherits broad access through repository secrets, environment variables, service connections and runtime tokens. In GitHub Actions, a workflow can expose credentials at build time even when the originating developer account only needed write access. That is why the runner is not just an execution environment; it is a temporary identity boundary where code, tokens and cloud trust converge. When long-lived secrets are present, the workflow becomes an efficient harvesting surface rather than a controlled automation step. See the CI/CD Pipeline Identity Security Guide.
Practical implication: Map every secret a runner can reach and treat that exposure path as a privileged access path, not a generic build dependency.
How workflow injection turns write access into secret theft
Workflow injection works because CI systems often trust repository changes to define execution. Once an attacker can place or modify a workflow file, they can run code inside the pipeline context and read secrets that are only supposed to exist at execution time. The danger grows when the workflow also has access to deployment environments or cached credentials, because the attacker does not need external malware to steal them. The compromise lives inside the pipeline definition itself. See the GitHub and GitHub Actions Security Guide.
Practical implication: Separate workflow authorship from workflow trust and require stronger controls for any path that can change runner behaviour.
Why OIDC federation still fails without claim binding
OIDC removes static cloud keys, but it does not remove trust decisions. If the cloud role accepts tokens from any branch, any workflow or any environment, an attacker can mint a valid token from a compromised repository context and exchange it for cloud access. The real control is not federation alone but constrained federation, where the token claims must match protected refs, approved workflows and explicit environments. Without that binding, the attacker simply swaps one credential type for another. See the Cloud Workload Identity Guide.
Practical implication: Bind federated identity to protected branches, environments and workflow claims before using OIDC as a replacement for static secrets.
Threat narrative
Attacker objective: The attacker wanted to harvest runtime secrets from CI/CD pipelines and convert those credentials into cloud and identity access before detection.
- Entry occurred through compromised GitHub accounts that could open probe pull requests or alter repository workflows. That gave the attacker a trusted path into CI/CD execution context rather than a direct cloud intrusion.
- Credential access followed when injected workflow steps dumped runtime secrets, environment variables and local credential files from the runner. The campaign also used OIDC minting and token replay to reach cloud and vault sessions.
- Escalation happened as the payload attempted trust-policy changes, federated credential injection and status-check forgery to survive rotation and branch protection. That turned a single workflow execution into a broader identity persistence problem.
- Impact was the rapid exfiltration of CI/CD secrets, cloud IAM credentials and developer tokens, followed by log deletion and commit reversion. The goal was to drain the pipeline before defenders could observe a stable forensic trail.
Breaches seen in the wild
- tj-actions/changed-files compromise 2025: A stolen bot token let attackers poison tj-actions/changed-files so pipelines printed their CI/CD secrets to public logs (CVE-2025-30066).
- CircleCI breach 2023: Malware stole a CircleCI engineer's SSO session; attackers exfiltrated customers' CI/CD secrets and keys, forcing a platform-wide rotation.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
Standing secrets in CI/CD create identity blast radius: The campaign worked because workflow execution could reach static credentials that should never have been available at build time. When a runner can read cloud keys, database passwords and AI tokens, the pipeline is effectively a privileged identity store. Practitioners need to stop treating secrets exposure as a side effect of automation and recognise it as a structural governance failure.
Workflow trust is the control surface, not the repository alone: A compromised account was only the entry point; the real weakness was that repository changes could alter trusted execution paths. That means branch protection, review gates and protected environments are part of identity governance for CI/CD, not just software release hygiene. CI/CD pipeline identity security is the natural control domain here.
Static secrets create persistence that survives rotation: The payload did not rely only on theft. It also tried to rewrite trust policies and federated credentials so access could outlive password or key rotation. That is a governance assumption collapse: defenders often assume revocation ends the event, but persistent trust artifacts can keep the identity alive after the original secret is gone. The implication is that offboarding logic must extend beyond the leaked credential itself.
Speed is now part of the threat model: The cleanup loop compressed intrusion, harvesting and erasure into roughly a minute in many cases. That makes detection windows shorter than many enterprise review cycles and turns delayed investigation into a structural disadvantage. Security teams should evaluate CI/CD identity controls against machine-paced abuse, not human incident timelines.
Secretless delivery is the architectural answer: Dynamic, ephemeral credentials reduce what a compromised workflow can steal, but they only work when the surrounding trust chain is tight. The point is not to add another secrets layer. It is to remove standing secrets from the pipeline and make identity proofing, claim validation and least privilege the default execution model.
From our research library:
- 96% of organisations store secrets outside of secrets managers in vulnerable locations including code, config files, and CI/CD tools, according to the Ultimate Guide to NHIs.
- Read next: CI/CD Pipeline Identity Security Guide
What this signals
Secretless CI/CD will move from preference to baseline: The campaign shows that static credentials in pipelines are now a predictable exploitation target, not an edge case. As more organisations tie cloud and AI access to build systems, the weakest point becomes the trust chain around workflow execution, not the secrets manager itself.
Pipeline governance will need to absorb identity controls: Branch protection, environment approval and claim validation are becoming part of identity assurance, because they determine whether a workflow is allowed to exchange identity for privilege. That shifts CI/CD from a delivery concern into a control plane for machine access.
Credential concentration is becoming a measurable risk: 96% of organisations store secrets outside of secrets managers in vulnerable locations including code, config files, and CI/CD tools, according to the Ultimate Guide to NHIs. That makes pipeline compromise a high-probability route to cloud and developer credential exposure.
For practitioners
- Revoke exposed developer access immediately Invalidate personal access tokens, OAuth grants and SSH keys tied to accounts that touched suspicious pull requests or workflow changes, then review any organization-level access those accounts inherited.
- Hunt for workflow-level indicators of compromise Search repositories for suspicious branch names, pull request titles, workflow file changes and missing sequential run numbers that indicate commit reversion and log deletion.
- Rotate every secret reachable from affected runners Replace cloud access keys, deployment tokens, AI API keys and any other credentials that the workflow could read, then assume the runner context itself was exposed.
- Bind federated cloud access to protected workflow claims Restrict OIDC trust to approved branches, environments and workflow references so a valid token from an untrusted path cannot exchange for cloud sessions.
- Remove standing credentials from CI/CD pipelines Move build-time access to short-lived, least-privilege credentials and deny runner permissions that can create keys, rewrite trust policies or persist federated access.
Key takeaways
- The core risk is not just a compromised GitHub account, but a pipeline architecture that lets one token expose many downstream credentials.
- The campaign combined workflow injection, secret harvesting and rapid cleanup, which compresses defender response time to minutes rather than hours.
- The structural fix is to remove standing secrets from CI/CD, constrain federated trust to protected claims and treat workflow governance as identity governance.
Related NHI Mgmt Group resources: Guide to the Secret Sprawl Challenge · CI/CD Pipeline Identity Security Guide · Cloud Workload Identity Guide · Top 10 CI/CD Security Risks
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Non-Human Identity Top 10 and MITRE ATT&CK address the attack and risk surface, while NIST SP 800-53 Rev 5 and NIST Zero Trust (SP 800-207) set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-02 — Secret Leakage | The campaign steals secrets from workflow context and runtime logs, which is direct secret leakage. |
| NHI-04 — Insecure Authentication | Stolen tokens and weak workflow trust let attackers authenticate from compromised GitHub contexts. | |
| NHI-07 — Long-Lived Secrets | The attack depends on standing keys and tokens that remain useful long enough to be harvested. | |
| Recommendation — Scan runner contexts and logs for exposed secrets, then revoke any leaked credentials immediately. Bind workflow authentication to protected claims and reject tokens from untrusted execution paths. Replace long-lived CI/CD secrets with short-lived credentials and rotate any persistent secrets out of pipelines. | ||
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | The incident is driven by leaked authenticators and the need to revoke and rotate them quickly. |
| Recommendation — Enforce authenticator rotation and revocation for every credential the pipeline can access. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | The attack chain centers on credential harvesting from runners and pivoting into cloud and vault access. |
| Recommendation — Map pipeline theft to credential access and lateral movement detections in your monitoring pipeline. | ||
| NIST Zero Trust (SP 800-207) | PDP/PEP — Policy Decision Point / Policy Enforcement Point | Protected branches, environments and claim checks act as enforcement points for workflow access decisions. |
| Recommendation — Use policy enforcement points to block untrusted workflow claims before they can mint cloud access. | ||
Key terms
- GitHub Action: A GitHub Action is an automated workflow step that runs inside GitHub-based development and delivery processes. It can be used for build, test, or maintenance tasks, but it also becomes part of the security boundary because compromised actions can access secrets and influence downstream systems.
- CI/CD runner: A CI/CD runner is the execution environment that performs build, test, or deployment jobs. It often has access to source code, tokens, and cloud credentials, which makes it a high-value identity surface when workflows are compromised.
- OIDC Federation: OIDC federation is a token-based trust model that lets one system accept identity assertions from another. It is commonly used to avoid static cross-environment credentials and to issue short-lived access based on trusted token claims. The control value depends on how tightly the trust relationships are governed.
- Static Secret: A secret, such as an API key or password, that does not change automatically over time. Static secrets require manual or scheduled rotation and represent a higher security risk than dynamic secrets or managed identities.
What's in the full article
Akeyless's full research covers the operational detail this post intentionally leaves for the source:
- The exact pull request, branch and workflow naming patterns defenders can use for repository hunting.
- The runner, cloud and identity log sources needed to confirm whether a pipeline was executed or cleaned up.
- The specific trust-policy and federated-credential artefacts attackers altered to survive rotation.
- The implementation details of secretless CI/CD controls across GitHub OIDC and dynamic secrets.
👉 Akeyless's full research covers the attack chain, cleanup evidence and pipeline containment steps.
Deepen your knowledge
NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.
Published by the NHIMG editorial team on November 16, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org