Trigger.dev EUVD-2026-107017

Requires the attacker to hold an UNRESTRICTED environment API key (scoped/restricted keys do not qualify) and to know or supply a victim's foreign parentRunId, then call POST /api/v3/batches or POST /api/v2/tasks/batch with the resumeParentOnCompletion option against a Trigger.dev instance before version 4.7.0. Limiting factors: exploitation is gated on possessing a valid unrestricted key (PR:L - not anonymous), and meaningful cross-tenant impact depends on a shared/multi-tenant deployment (e.g., Trigger.dev Cloud or a shared self-hosted instance) where runs from other orgs or production projects are reachable; single-tenant isolated deployments materially reduce real-world risk.

This is a moderate, authenticated-access authorization flaw (CWE-639, IDOR-style) rather than a critical remote-takeover bug. The supplied CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N) confirms network-reachable, low-complexity exploitation but requires an attacker to already hold a valid unrestricted environment API key (PR:L) - so this is not an unauthenticated internet-facing bug. Impact is scoped to integrity and availability of task runs (VI:H/VA:H) with no confidentiality impact (VC:N), meaning an attacker can disrupt/poison other tenants' runs but cannot read their data. No EPSS score, no CISA KEV listing, and no public POC were provided in the input, so there is no signal of active or widespread exploitation; the primary risk driver is the cross-organization/multi-tenant blast radius, which matters most for the hosted Trigger.dev Cloud or shared self-hosted deployments. One caveat to verify: the description emphasizes 'cross-organization' impact, which could arguably justify a CVSS scope change, yet the vendor's 4.0 vector sets all subsequent-system metrics to N - treat the tenancy-crossing severity with that ambiguity in mind.

An attacker who possesses any unrestricted environment API key (e.g., a developer with legitimate access to one project, or a leaked key) sends a crafted POST to /api/v3/batches or /api/v2/tasks/batch referencing a parentRunId belonging to a different, higher-value production or cross-organization run. By setting resumeParentOnCompletion, the attacker ties that victim run's lifecycle to attacker-controlled batch tasks, allowing them to indefinitely suspend, block, or inject errors into runs they do not own. No public exploit identified at time of analysis, but the technique requires only standard authenticated API calls and is low-complexity.