@zuke/gh EUVD-2026-114134

Exploitation requires a very specific set of preconditions, all drawn from the advisory: (1) a build that gates a release on @zuke/gh githubWorkflow(...) running in marker (display-title correlation) mode; (2) the dispatched workflow must live in a separate repository whose workflow_dispatch / actions:write permission is held by a party the releasing repository does not equally trust - e.g. a shared org workflow or any repo with more contributors than the one cutting the release; and (3) the attacker must dispatch the same workflow on a ref they control during the gate's wait window. Reading the marker itself needs no special access because it is echoed into a run title visible to anyone who can list the workflow's runs. Limiting factors: this does not affect single-trust-domain setups where actions:write holders are exactly the trusted releasers, does not affect the best-effort mode in the same way, and is closed by upgrading to 1.10.0; branch protection does NOT mitigate it because workflow_dispatch is not gated by branch protection. Affected versions: @zuke/gh < 1.10.0.

CVSS 3.1 base score is 7.1 (High) with vector AV:N/AC:H/PR:L/UI:N/S:C/C:N/I:H/A:L, meaning network-reachable but high attack complexity, requiring some privilege, and a scope change - integrity (I:H) is the dominant impact with a secondary availability impact (A:L) and no confidentiality impact (C:N). There is no CISA KEV listing and no EPSS score in the provided data, so exploitation probability cannot be quantified; the status is no public exploit identified at time of analysis (the four references are the GHSA advisory, PR #542, the fix commit, and the release tag - all patch artifacts, not exploit code). The vuln.today decision label is this_week, indicating elevated but not emergency urgency, which is consistent with a high-integrity-impact flaw whose exploitation is gated behind a specific non-default trust configuration (AC:H). The real-world priority is conditional rather than universal: it is a genuine integrity threat to release pipelines only where @zuke/gh githubWorkflow(...) runs in marker mode AND the dispatched workflow lives in a repository whose workflow_dispatch permission is held by a less-trusted party. One signal conflicts with the data: the 'Information Disclosure' tag contradicts the CVSS vector (C:N) and CWE-345 - the actual impact is authenticity/integrity of the gated release, not disclosure, so that tag should be disregarded.

An attacker who holds workflow_dispatch on a shared or multi-contributor workflow repository reads the display-title marker from that workflow's publicly visible run list, then dispatches the same workflow on a branch they control so the gate adopts the attacker's run. A passing substituted run authorizes a release whose tested code never actually ran, while deliberately failing or cancelling the run blocks a legitimate release. No public exploit code is identified; the mechanism is described in detail in the GHSA advisory, and the high attack complexity reflects the required cross-trust-domain configuration rather than a difficult trigger.