Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-114070 Overview
CVE-2026-114070 is an information disclosure vulnerability in Trigger.dev versions 4.5.0 through 4.6.0. The Dashboard Agent input proxy stores a delegated user-actor token inside browser-readable session input records. An attacker who holds a chat's public session token can query the in records endpoint and extract the delegated token from payload.metadata.userActorToken. The recovered token can then be replayed against the environment JWT exchange endpoint, giving the attacker access tied to the delegated user actor. The weakness is classified under CWE-522: Insufficiently Protected Credentials.
Critical Impact
An attacker with a valid chat public session token can recover a delegated user-actor token and exchange it for an environment JWT, enabling unauthorized access to protected Trigger.dev environment APIs.
Affected Products
- Trigger.dev versions 4.5.0 up to (but not including) 4.6.0
- Trigger.dev Dashboard Agent input proxy component
- Deployments exposing chat public session tokens to untrusted browser contexts
Discovery Timeline
- 2026-10-21 - CVE-2026-114070 published to NVD
- 2026-10-21 - Last updated in NVD database
Technical Details for CVE-2026-114070
Vulnerability Analysis
Trigger.dev's Dashboard Agent exposes an input proxy that persists session input records for client-side rendering. The proxy writes a delegated user-actor token into the record payload under payload.metadata.userActorToken. Because the input records endpoint (in) is reachable with only the chat's public session token, any party holding that session token can read the stored payload and extract the embedded credential. Credentials intended for server-to-server use are therefore exposed to any browser or script that can present the session token.
Root Cause
The root cause is improper credential storage in a client-reachable record. The delegated user-actor token should remain confined to server-side state or a short-lived, scoped exchange. Instead, the token is serialized into the session input record alongside non-sensitive metadata. Access control on the in records endpoint is bound to the public session token, which does not provide the authorization level required to protect a delegated actor credential. This design pattern maps directly to [CWE-522] (Insufficiently Protected Credentials).
Attack Vector
Exploitation is network-based and requires that the attacker already possess a chat's public session token, which can be shared through chat links, logs, or shoulder-surfed URLs. The attacker calls the input records endpoint, parses the returned JSON payload, and reads payload.metadata.userActorToken. The token is then submitted to the environment JWT exchange endpoint to obtain an environment-scoped JWT. With that JWT, the attacker can invoke APIs under the delegated user actor's identity until the token expires or is revoked. No special tooling or memory corruption is required; standard HTTP tooling is sufficient. Technical specifics are documented in the GitHub Security Advisory GHSA-wpp9-mq73-fhqm and the VulnCheck advisory on token disclosure.
Detection Methods for CVE-2026-114070
Indicators of Compromise
- Requests to the Dashboard Agent in records endpoint from unexpected IP addresses or user agents.
- JWT exchange calls submitting a userActorToken immediately after a session input record read.
- Session input records whose stored payload contains a metadata.userActorToken field.
- Environment JWTs issued to actors that did not initiate interactive dashboard sessions.
Detection Strategies
- Inspect application logs for sequential access patterns: in record fetch followed by an environment JWT exchange from the same client.
- Audit stored session input records for any payload containing userActorToken and treat matches as exposed credentials.
- Correlate chat public session token usage with API calls that assume delegated actor identity.
Monitoring Recommendations
- Alert on high-volume reads of the Dashboard Agent input records endpoint from a single session token.
- Track newly issued environment JWTs and compare the issuing actor against expected dashboard activity.
- Ingest Trigger.dev application and reverse-proxy logs into a centralized analytics platform to retain evidence for investigation.
How to Mitigate CVE-2026-114070
Immediate Actions Required
- Upgrade Trigger.dev to version 4.6.0 or later, which removes the token from browser-readable session input records.
- Rotate any delegated user-actor tokens and environment JWTs that may have been exposed while running an affected version.
- Invalidate active chat public session tokens that were in use during the vulnerable window.
- Review access logs for the in records endpoint and the environment JWT exchange endpoint for signs of abuse.
Patch Information
The maintainers released a fix in Trigger.dev 4.6.0. Review the remediation guidance in the GitHub Security Advisory GHSA-wpp9-mq73-fhqm and the corresponding VulnCheck advisory before deployment. Confirm that upgraded builds no longer persist userActorToken inside session input payloads.
Workarounds
- Restrict access to the Dashboard Agent in records endpoint at the reverse proxy or WAF until the upgrade is complete.
- Shorten the lifetime of chat public session tokens and delegated user-actor tokens to limit the replay window.
- Scrub payload.metadata.userActorToken fields from any stored or cached session input records.
- Treat chat public session tokens as sensitive and avoid logging or sharing them outside trusted channels.
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

