Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

NewsIncident

An official OpenAI skill told apps to send tokens to a domain that did not exist; researchers registered it and the tokens came

A skill in OpenAI's official GitHub skills repository, for deploying applications on Render, instructed production apps to send their tokens to render-internal.com, a domain nobody had registered, CSO Online reported on October 16. AIR Security's Research Lab found Codex putting the domain into real application code as the default, registered it, pointed it at a passive sinkhole and began receiving tokens and API calls from real production apps. The mistake dated from a January commit; after AIR's report it was resolved in mid-August, with Render removing every reference, OpenAI purging the Render plugin from its plugins repository and the domain handed to Render. How many apps were affected was not published. AIR's Niv Hoffman: vet and monitor skills like any other supply-chain component. CSO also notes a trojanized skill set typosquatting Paperclip and Browser Use on skills.sh that reached 1.7 million installations.

Why it matters

A skill is code your agent runs, and an official source is a reason to review it rather than skip the review. Pin the skills you use, read generated code for hosts you do not recognize, and alert on outbound calls to domains that are not yours.

Sources

Read it with the rest of the news in the Learning Center