@atlasauth/embed@0.2.2
AI Security Review
scanned 21d ago · by lpm-firewall-aiThe source shows a browser authentication widget whose login, social, and passkey flows use the configured Atlas API, with no install hook or verified account-setting mutation. The unusual behavior is device and interaction telemetry sent to that configured API during authentication flows.
Decision evidence
public snapshot- package.json points to the browser widget entrypoint and lists only build, typecheck, and test scripts; it has no install lifecycle hook.
- The inspected runtime call sites handle user-facing sign-in, sign-up, password reset, social login, and passkey flows through the widget or Atlas client.
- Telemetry contains a device identifier, device fingerprint, browser characteristics, and interaction metrics, and is sent to the configured frontend API during authentication flows.
- The frontend API is read from widget configuration and is used as the Atlas client base URL.
- The widget sends device and interaction telemetry during authentication flows; this is privacy-relevant collection, but the inspected source routes it to the configured frontend API and does not include entered credentials in the telemetry payload.
Source & flagged code
8 flaggedBrowser source reuses an authenticated session to collect identity data and mutate account settings while reporting externally.
dist/embed.jsView on unpkg · L1A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.
dist/widget/social.js#virtual:normalized:round1View on unpkgFindings
Affected versions and remediation
This report applies to @atlasauth/embed@0.2.2.
See all versions of @atlasauth/embed for other recorded verdicts.
- Review the evidence and your use of @atlasauth/embed@0.2.2 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.