The images in this article were generated with artificial intelligence. How we publish
A security researcher published exploit code for a zero-day vulnerability in Visual Studio Code that allows to steal GitHub authentication tokens with a single click on a malicious link. According to the author's public analysis, the failure abuses the web-view system and the passing of messages in VS Code to run JavaScript within a drunk view, simulate key pulsations and force the installation of a malicious extension that captures the token OAuth that github.dev receives from github.com.
Understanding the technique is key to risk assessment: github.dev performs a POST with an OAuth token which, according to the researcher, is not limited to the specific repository, but provides access to all the repositories to which the user can access. That means that an attacker who gets that token can consult the GitHub API to list and clone private repositories, pivote or extract secrets stored in code. Researcher Ammar Askar published an analysis and the proof-of-concept on his blog and on GitHub showing the chain of exploitation; the technical detail can be read in his article and in the repository with the PoC: Askar's explanation and PoC in GitHub.

From the vulnerability classification perspective, Microsoft and other suppliers consider a defect as zero-day when it is publicly disclosed or actively exploited without an official patch; Microsoft documents this criterion in its vulnerability management guide: What a zero- day is. In this case the failure had no CVE or patch at the time of publication, so the community should take immediate precautions.
For users and development teams, the practical implications are direct: a single click on a link that opens github.dev under certain conditions could result in commitment of tokens with wide access. The tokens credentials are equivalent to access keys: their exposure allows actions on behalf of the user and may affect private repositories, automated integrations and deployments that depend on such credentials.
Immediate measures that reduce the risk while there is no patch: close session on github.dev and delete cookies and local site data for github.dev in the browser, so that the extension authorization dialog appears again (this requires manual confirmation before the extension receives a token). Avoid clicking on unverified links that open VS Code or github.dev from emails, chats or websites; prefer to open repositories from GitHub's web interface directly and, if you should use github.dev, do so from an isolated browser profile.
If you suspect commitment, mediation actions should include revoking any affected personal token, creating smaller-range tokens (using "fine-grained" personal access tokens where possible) and rotating credentials used by scripts or CI / CD. Check the API activity and audit records in your account or organization to detect suspicious calls and access repositories. GitHub offers guides on how to keep accounts and tokens safe; see your documentation for specific procedures on revocation and audit.

At the organizational level, the principle of minor privilege should be applied: to replace tokens with global permits with necessary minimum permits, to use mandatory multifactor authentication (MFA), to manage access through SSO or identity providers and to audit the OAuth apps and tokens authorized in the equipment accounts. It is also recommended to segregate development environments, limit the presence of secrets in repositories and monitor integrations that can be abused if a partner is committed.
This incident also reopens the discussion on trust and responsible disclosure processes: the investigator claims that his decision to publish immediately was motivated by previous experiences with the handling of reports by official channels. The tension between public disclosure and coordination with suppliers is recurrent and has effects on collective security; however, until there is a patch, the responsibility lies with users and managers to mitigate impact.
In short, treat this event as a reminder that tokens matter as much as passwords: maintain good credentials management practices, delete session data from github.dev if you do not trust the link you will open, revoke and rotate tokens when there is suspicion and apply stricter access controls in organizations. To expand how to protect your account and manage tokens, see GitHub's official documentation on security and authentication.
Related
More news on the same subject.

FBI and six countries link Integrity Technology Group to entity post theft in SE Asia
On October 8, the FBI and agencies in six countries issued a joint warning that assigns to a Chinese company, Integrity Technology Group, a sustained series of intrusions whose ...

Campaign with LLM and ARTEX attacks South Korean financial institutions and exfilters data
Security researchers have documented a campaign directed against South Korean financial institutions using language-driven attack tools to automate intrusions and data extractio...

ChainDrop campaign exposes tensorlake in npm; version 0.5.144 withdrawal
A package of npm called tensorlake, an SDK in TypeScript oriented to Tensorlake applications and services, was engaged in a supply chain campaign linked to the attack family kno...

Google reports DNS kidnapping: TLS certificates for google.com.gh, google.sl and google.as
Google reported on October 6 that attackers managed to issue unauthorized HTTPS certificates for Google and YouTube names after compromising authoritative DNS records of three t...

Cyber risk in 2026 moves to workflows and IA, according to Voice of the CISO
The data added by five editions of the Voice of the CISO study - including the most recent findings of 2026 - draw a less intense change than risk location: the threat is moving...

Phishing BitB points to advertising professionals and account managers to steal MFA
Security researchers have described a phishing campaign for advertising professionals and account managers that uses a human-operated platform to mimic ad products linked to IA ...

LibreOffice / OpenOffice Calc allows remote source execution when opening ODB / JDBC leaves
Researchers have shown that a malicious spreadsheet can force LibreOffice and Apache OpenOffice to run code controlled by an attacker at the time the file is opened, without sho...