@nalvietnam/avatar-cli@3.29.0
AI harness CLI for NAL Vietnam engineering
AI Security Review
scanned 15d ago · by lpm-firewall-aiThe doctor probe forwards an existing token to an operator-configurable endpoint without a visible destination restriction, creating conditional credential exposure. Because the probe is user-invoked and can be disabled, this warrants a warning rather than a publish block.
Decision evidence
public snapshot- The doctor command runs the telemetry probe by default; users can disable it with `--no-probe`.
- The probe reads an existing token and accepts `AVATAR_TELEMETRY_URL` as its destination without an evident host restriction.
- The probe sends that token in a bearer-authenticated POST, even though the body contains only a probe marker.
- The bundle comments claim the response can confirm token validity; this is source text, not proof that the endpoint is safe.
- The manifest's postinstall hook imports a welcome-screen module, while that module's body was not independently inspected.
- The bundle's agent configuration marker identifies a Claude workspace config path; this excerpt alone shows no write or mutation.
- The probe is part of the explicit doctor command and offers a `--no-probe` opt-out; no install-time probe was established.
- The configured default telemetry host belongs to the package's vendor domain.
- The cited agent configuration code only checks whether the Claude config path exists.
Source & flagged code
9 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgFindings
Affected versions and remediation
This report applies to @nalvietnam/avatar-cli@3.29.0.
See all versions of @nalvietnam/avatar-cli for other recorded verdicts.
- Review the evidence and your use of @nalvietnam/avatar-cli@3.29.0 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.