MalwarePublished 29 Jun 2026Updated 9 Jul 2026

openai-agents-helpers (npm) · Malicious Code

Affects openai-agents-helpersSource GCVE Database

Malicious package : remove openai-agents-helpers and rotate any secrets on hosts that installed it.

Data as of 10 Jul 2026, 09:29 UTC

Classification

Malicious package

Published to npm · OSV MAL-2026-6582

What to do

Remove openai-agents-helpers from every project, lockfile and image that pulls it in. Treat any host that installed it as compromised: rotate the credentials, tokens and keys it could read, from a clean machine.

Affected software

Product	Affected	Fixed in
npmopenai-agents-helpers	Affected =1.3.2	Fixed in–

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

Description

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (768ddf82a7644f1678a27f3037c8a3fc4fae5f00b6181d6507a49892d20a7fda) On `npm install`, scripts/postinstall.js automatically reads a broad set of installer-side identity and cloud-configuration files — ~/.gitconfig and the parent project's.git/config plus.git/logs/HEAD (committer emails), ~/.config/gh/hosts.yml (GitHub login), every ~/.ssh/*.pub file (key identity comments / emails), ~/.config/gcloud/properties (GCP project and account), ~/.aws/config (profile names and SSO identifiers), /etc/resolv.conf (corporate DNS search domains), os.hostname(), os.userInfo().username, current working directory, and the parent project's package.json. The collected data is bundled into a JSON payload and POSTed via https.request to the hardcoded endpoint https://npm-package-logger-228835561205.europe-west1.run.app/. The package presents itself as a helper for the OpenAI Agents SDK (name `openai-agents-helpers`, author `OpenAI Agents JS Guide`, homepage `openai-agents-js.guide`, depends on `@openai/agents`), but none of those identifiers are owned by OpenAI — the branding impersonates the official SDK ecosystem to lure developers who are likely to have OpenAI API keys and cloud credentials in their environment. Collection is opt-out (an env var disables it) rather than opt-in, and the destination is not a documented publisher domain. Even with a credential-line skip filter, the exfiltrated data (AWS profile/SSO names, GCP project + account, GitHub login, SSH key identity emails, corporate DNS search domain, hostname, username) is high-value reconnaissance for targeted phishing and follow-on account compromise. ## Source: ghsa-malware (6303be8ee54533f33251590555db82063d5274b4d11a626603cca87c849e7807) Any computer that has this package installed or running should be considered fully compromised. All secrets and keys stored on that computer should be rotated immediately from a different computer. The package should be removed, but as full control of the computer may have been given to an outside entity, there is no guarantee that removing the package will remove all malicious software resulting from installing it.

Impact

The malicious package compromises the confidentiality of sensitive credentials and identity data on affected systems. This exposure facilitates targeted phishing attacks and account takeover attempts by adversaries. Because the package executes automatically on installation and accesses secret-adjacent files, any system with this package installed should be treated as fully compromised. Attackers may have persistent access beyond the presence of the package itself, increasing remediation complexity and risk.

Defensive guidance

Immediate removal of the 'openai-agents-helpers' package version 1.3.2 is recommended. All secrets, keys, and credentials stored on affected systems should be rotated immediately from a separate, uncompromised environment. Treat any system with this package installed as fully compromised and conduct a thorough security assessment. Monitor for any unauthorized access or suspicious activity related to exposed credentials. Exercise caution when installing packages that impersonate trusted vendors and verify package authenticity before use.

Summary

The 'openai-agents-helpers' npm package version 1.3.2 includes a post-install script that harvests a wide range of sensitive information from the host environment, such as SSH public keys (including email comments), AWS and GCP credentials and profiles, GitHub CLI user data, git commit metadata, hostname, username, and CI provider details. This data is sent via HTTPS POST to a hardcoded external endpoint unrelated to OpenAI. The package falsely presents itself as an official OpenAI tool, misleading users. The exfiltrated data enables high-value reconnaissance for attackers aiming at phishing and account takeover. The malicious code triggers automatically on installation without opt-in, and the package accesses directories and files unrelated to its legitimate function. Removal of the package does not guarantee full remediation due to potential further compromise.

References

Community reviews

Loading reviews…

What to do

Affected software

Description

AI analysisMachine-generated, updated 8 Jul 2026, 15:06 UTC

Technical details

References

Community reviews