Vulnerability  ·  2026-08-14

Trigger.dev AI agent/workflow platform: critical background-worker RCE and metadata-path injection

VulnerabilityHigh impactGlobal
Trigger.dev, a platform for building and deploying managed AI agents and workflows, contains CVE-2026-73656 (CVSS 9.9, Critical) where POST /api/v1/deployments/:deploymentId/background-workers passes insufficiently validated input into CreateDeploymentBackgroundWorkerServiceV4.call(), and CVE-2026-73654 (CVSS 8.5, High) where the PUT /api/v1/runs/:runId/metadata endpoint passes attacker-controlled operation.key values into new JSONHeroPath(operation.key).set(newMetadata, value), allowing path-based injection into run metadata.
Trigger.dev orchestrates AI agent execution and background workers; a near-maximum-severity flaw in the worker-creation endpoint of an AI agent deployment platform represents a direct path to compromising the execution environment for hosted AI agent workloads, with a large blast radius across any organization using Trigger.dev to run agentic pipelines.
Unauthenticated/authenticated API request to the background-worker creation endpoint or the run-metadata update endpoint, injecting attacker-controlled values that are processed without adequate validation
Trigger.dev < 4.5.6
Upgrade to Trigger.dev 4.5.6 or later per vendor commit fixes
NVD - CVE-2026-73656GitHub commit reference
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →