Trigger.dev EUVD-2026-106888

Exploitation requires: (1) adjacent network access to the supervisor workload server (AV:A - not routable from the open internet by default); (2) possession of a VALID run/snapshot pair, i.e. an existing legitimate low-privilege workload credential (PR:L); and (3) issuing requests to the specific attempts/start endpoint of the supervisor workload API. The core defect is the absence of any ownership/tenant verification tying the run/snapshot to its deployment. Limiting factors that reduce real-world risk: the adjacency requirement, the need for a valid run/snapshot pair rather than anonymous access, and the elevated attack complexity/attack requirement (AC:H/AT:P) indicating the attacker must satisfy specific preconditions such as a usable snapshot pairing. Greatest exposure is in multi-tenant or hosted Trigger.dev installations where untrusted workloads share a supervisor; single-tenant self-hosted deployments with no untrusted workloads have materially lower exposure.

This is a credible mid-to-high priority issue for Trigger.dev operators, but not a mass-exploitation emergency. The CVSS 4.0 vector (AV:A/AC:H/AT:P/PR:L) describes an attacker who already holds a legitimate low-privilege foothold (a valid run/snapshot pair, PR:L) and adjacent network access to the supervisor workload server (AV:A), with meaningful attack complexity and a present attack requirement (AC:H/AT:P) - not a trivial internet-wide unauthenticated bug despite the CWE-306 'missing authentication' label. The high impact (VC:H/VI:H and scope/subsequent-system impact SC:H/SI:H) reflects genuine cross-tenant data exposure (resolved environment variables, run JWTs) and run manipulation. No EPSS score was supplied, the CVE is NOT in CISA KEV, and no POC is referenced, so there is no public exploit identified at time of analysis and no evidence of active exploitation. The practical picture is a serious multi-tenant isolation failure that matters most in shared/hosted Trigger.dev deployments, but the adjacency requirement, need for a valid run/snapshot pair, and elevated complexity all constrain real-world reach. Note the CWE-306 'missing authentication' framing is in tension with PR:L in the vector: the flaw is really a missing OWNERSHIP/authorization check on an endpoint that still requires some valid workload credential - verify with the vendor advisory (GHSA-jc26-22qp-cgqj) before treating it as fully unauthenticated.

In a shared Trigger.dev deployment, a tenant who already runs a legitimate workload obtains a valid run/snapshot pair and, from an adjacent network position, calls the supervisor workload server's attempts/start endpoint referencing another tenant's run. Because ownership is never verified, the server returns that victim tenant's resolved environment variables and run JWTs and lets the attacker start, complete, or suspend the victim's runs. No public exploit code is referenced, so exploitation would require custom tooling, and the AC:H/AT:P metrics imply non-trivial conditions such as guessing or obtaining a valid snapshot pairing.