Write access was tested before the malicious change

Aikido Security reports a campaign that used compromised GitHub accounts to probe repositories with a wfx-docs branch and a routine-looking documentation change. In repositories where the account could write, the attacker then pushed changes intended to run in CI and collect secrets available to the workflow.

The researchers estimate thousands of accounts and repositories were involved. Those totals are an evolving campaign measurement, not a count of organizations with confirmed stolen credentials. Public GitHub workflow history shows the reported branch and run pattern in at least one repository, but that example does not establish the full campaign size.

A revert is not a recovery step

Removing the malicious file or branch stops future runs. It does not take back secrets that a completed workflow could already read.

Start with the repository events, then follow the secrets

  1. 1

    Search the reported pattern. Review recent write events, pull requests and workflow runs for wfx-docs, docs: dependency audit notes and unexplained CI helper changes from 10 November onward.

  2. 2

    Preserve the history. Capture commits, run IDs, logs and audit events before cleanup. A deleted branch or run can make the sequence harder to reconstruct.

  3. 3

    Scope credential access. For any workflow that ran attacker-controlled code, identify repository secrets, cloud credentials, tokens and OIDC permissions it could reach. Revoke or rotate those credentials and review their use.

Judge impact from the workflow's permissions

The most useful question is whether the malicious step executed with access to a secret or trusted identity. Review branch protections, workflow triggers and token scopes alongside the timeline. A probe in a repository and an executed secret-stealing workflow have different implications, so report them separately.