/

CVE-2026-73659

Severity:

HIGH RISK

(8.1)

Path Traversal (CWE-22) in Trigger.dev Object Store

CVE-2026-73659

Deferred

Vulnerability type

Category:

Path Traversal

CWE:

CWE-22

Impact:

Cross-Organization Data Disclosure and Overwrite

Severity:

HIGH RISK

(8.1)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

Trigger.dev packet presign routes accept a caller-controlled filename without adequately restricting path traversal sequences. On multi-organization self-hosted instances, a project API key may therefore read or overwrite another organization's offloaded task payloads and outputs.

Affected context

The issue occurs in Trigger.dev deployments using the affected packet presign routes and object-store integration. It requires a filename to influence presigned URL or request generation without being securely confined to the intended packets/<projectRef>/<env>/ storage prefix; the risk is especially significant in self-hosted environments serving multiple organizations.

Exploit preconditions

  • Attack surface: Packet presign API routes in the web application
  • Required access: A valid project API key or other access capable of invoking the presign functionality
  • Trigger point: Processing a caller-controlled filename while generating a presigned URL or request
  • Why exploitation is possible: Path traversal components can cause storage-key resolution to escape the project and environment prefix, potentially reaching objects belonging to another organization

Why this happens

The vulnerability results from treating a caller-controlled filename as a trusted object-store path component. When traversal elements are passed into storage-key construction without canonicalization and confinement checks, the resulting key may resolve outside the directory-like prefix intended for the request. This behavior maps to CWE-22: Improper Limitation of a Pathname to a Restricted Directory, applied here to object-store key generation and authorization boundaries.

Potential impact

  • An attacker may read another organization's offloaded task payloads or outputs
  • An attacker may overwrite objects belonging to another organization
  • Confidential workflow data could be disclosed
  • Stored task results could be tampered with or replaced
  • Cross-tenant isolation could be weakened on shared self-hosted instances

General mitigation guidance

Mitigation generally focuses on treating filenames as untrusted identifiers and enforcing storage-boundary checks before presigned URLs or requests are created.

Common defensive patterns include:

  • Rejecting traversal components and unsafe path syntax
  • Canonicalizing candidate object keys before authorization decisions
  • Verifying that resolved keys remain within the intended project and environment prefix
  • Deriving storage paths from trusted server-side identifiers rather than caller-provided paths
  • Binding authorization checks to the project, organization, environment, and object being accessed
  • Using separate storage namespaces or credentials where practical to limit cross-organization access
  • Applying least-privilege permissions to object-store credentials
  • Avoiding storage APIs that implicitly normalize paths without an explicit containment check

How teams detect issues like this

High-level detection can include:

  • Static analysis for concatenation or interpolation of request-controlled filenames into object-store keys
  • Code review checks for missing canonicalization, containment validation, or tenant-bound authorization
  • Tests covering traversal markers, encoded separators, absolute-path forms, and normalization edge cases
  • Policy checks requiring server-derived tenant and environment prefixes for presigned operations
  • Audit logging for presign requests, resolved object keys, project identifiers, and organization boundaries
  • Monitoring for unusual cross-project object access or repeated presign requests involving malformed filenames

Automated security review tools may help identify untrusted path construction, but findings should be validated against the storage provider's key-normalization behavior.

Sources

SourceLink
github.comgithub.com/triggerdotdev/trigger.dev/commit/db4074df54db06b0656becf3f974345c90fa202e
github.comgithub.com/triggerdotdev/trigger.dev/pull/3830
github.comgithub.com/triggerdotdev/trigger.dev/releases/tag/v4.5.0
github.comgithub.com/triggerdotdev/trigger.dev/security/advisories/GHSA-m3mf-37q7-8928

Frequently asked

questions

It can be exploitable through the packet presign API when the attacker has network access to the service and a project API key or equivalent permission to invoke the affected routes.

Yes. The attacker generally needs valid application credentials or another authorized route to the presign functionality; anonymous access is not implied.

Treat filenames as untrusted input, generate keys from trusted tenant-bound identifiers, and verify after normalization that every key remains within the authorized project and environment prefix.

Combine strict object-key validation with tenant-aware authorization, least-privilege storage credentials, isolated namespaces where appropriate, and automated tests for path traversal and normalization edge cases.

Find out what's already
exploitable in your codebase.