On 16 November 2026, GitHub's Advisory Database reviewed @xzvbailss/baileys
as malware. The package is a fork of Baileys, the WebSocket
library most WhatsApp automation tools in Node.js are built on. The fork does what Baileys does —
and then, without telling the user, subscribes their WhatsApp account to a channel the publisher
picked.

The advisory credits OpenSSF's malicious-packages project, record MAL-2026-19946.

What the real package looks like next to the fork

The comparison is the whole story. baileys (no scope) has existed since September 2022, has gone
through 64 published versions, is maintained under WhiskeySockets/Baileys
on GitHub, and its registry.npmjs.org record still lists that repository URL.

@xzvbailss/baileys was created on 11 November 2026 — five days before the advisory. Its
package.json repository field doesn't point to GitHub at all; it points to a Telegram handle,
https://t.me/JustRxVz. Its description is refreshingly candid for a malicious package: "hello I
am Rz, I am a student trying to develop a WhatsApp library via baileys and this is my baileys
xzvbailss." The same maintainer account, justrxvz, also owns an unscoped package called
xzvbailss — cited in the advisory as a separate malware record, MAL-2026-18702.

What it does, by version

The advisory describes two behaviours, both triggered after a delay so they don't fire during a
quick test install:

  • Versions 1.0.0 and 5.0.0 (both published 11 November, a minute apart): a hardcoded channel ID triggers a FOLLOW query 2.5 minutes after the socket connects, retried every 5 seconds. Errors are swallowed silently.
  • Version 6.0.0 (published a few hours later the same day): the channel ID is now base64-encoded and decoded at load. The follow runs on every connection.update "open" event, after a 90-second delay, with failures logged only at debug level — quieter, and harder to spot by reading the diff.

The advisory lists only those three versions as affected. The registry shows six more releases
after them — 6.0.1 through 6.0.9, all published within a few hours of each other on 13
November, two days after the advisory's sample and three days before it was reviewed.

We checked the version that's live right now

So we pulled the tarball for 6.0.9, the version npm install @xzvbailss/baileys would install
today, and looked for the behaviour ourselves. It's in lib/Socket/newsletter.js:

// XzV Newsletter Auto-Follow: follow one configured newsletter after connection. const newsletterAutoFollowId = config.newsletterAutoFollowId ?? '120363430764800148@newsletter'; const AUTO_FOLLOW_DELAY = 10_000; const AUTO_FOLLOW_RETRY_DELAY = 20_000; const AUTO_FOLLOW_MAX_ATTEMPTS = 3; 

No base64 this time — the comment names exactly what the code does. A timer fires 10 seconds after
the socket opens, calls the same FOLLOW query ID the advisory describes, retries up to three
times 20 seconds apart, and skips itself quietly if the configured ID isn't a well-formed
newsletter JID. The hardcoded channel differs from the one the advisory quotes for 6.0.0, which
tells us the operator changed the target at least once across the six patch releases — not that the
behaviour ever went away.

In short: GitHub's "affected versions" field is accurate for the versions it reviewed, but the
package kept shipping the same feature after that review window, and nothing currently stops
npm install @xzvbailss/baileys from pulling it.

The family this sits in

The advisory ties @xzvbailss/baileys to two other records: MAL-2026-17342, a "PhantomSub"
pattern it says matches the same behaviour, and MAL-2026-18873, a different package,
@poucode/baileys, that the advisory says shares a byte-identical copy of the same luxu.js file.
We checked: @poucode/baileys exists, was created on 11 October 2026 — a month before the
xzvbailss fork — and is also still resolvable, currently at version 2.0.1. We didn't go further
and compare the two packages' files ourselves; that claim is the advisory's, not something we
re-verified byte for byte.

Where this fits on the pull request

There's no typo here and no install script. @xzvbailss/baileys isn't trying to look like
baileys by misspelling it — it's an entirely different, scoped package name, and nobody adds it
to a project by accident. The signal a check can act on isn't the name; it's that the package is
five days old with no history, forked from something 64 releases and four years deep, pointing at a
Telegram handle instead of a repository. That's exactly the "newly published, low-adoption
dependency" pattern MagAudit's check is built to flag on the pull request that adds the dependency
line — whether that stops the merge depends on whether the team has made the check required in
their own branch protection settings, the same as any other check on the PR. A general-purpose AI
reviewer reading "@xzvbailss/baileys": "^6.0.9" in a diff has nothing in the line itself to go on;
the age and the fork relationship only show up by querying the registry, which is what a dedicated
check is there to do.

What this doesn't tell you

The advisory doesn't say what the followed channel publishes, how many accounts the auto-follow has
actually reached, or whether justrxvz controls the channel directly or is running this for someone
else. We don't know why the operator moved from a silent 2.5-minute delay to a candidly-commented
10-second one — confidence that reviewers wouldn't catch it, indifference, or simply sloppy
iteration. We didn't check versions 5.0.0 through 6.0.8 individually; our own verification
covers the advisory's description of 1.0.0/5.0.0/6.0.0 and our own read of 6.0.9, nothing in
between. And we took the advisory's word for the shared luxu.js file with @poucode/baileys
rather than diffing the two packages ourselves.

Disclosure: this article was produced automatically by our software agents (an AI model or a report template) from measurements they ran themselves, with no human editorial review before publication (EU AI Act, art. 50). Every figure is re-checked daily against its live source and corrected here if it drifts.