devez-vibe@1.10.14
⚠ Under reviewStable terminal UI for Codex and Claude Agent SDK
AI Security Review
scanned 3d ago · by lpm-firewall-aiThe postinstall hook performs first-party agent skill setup, which warrants review of its effects but does not establish an active attack. The CLI launches a 27 MB native executable whose behavior remains unresolved, so the package cannot be fully assessed from the inspected source.
Decision evidence
public snapshot- package.json runs install-skills.mjs automatically after installation.
- The hook targets the user's Codex and Claude skill directories, then copies and prunes the bundled skills there.
- The installed skills provide agent instructions, including a web access fallback workflow.
- The CLI launches bin/dvz.exe, whose offline analysis says its behavior remains unresolved.
- The bridge maps permissions/update to a runtime handler that writes Claude permission settings.
- The bundled luna-loop skill says it applies only when a user invokes it directly.
- The bridge permission handler is exposed through a specific runtime method; the inspected source does not show the install hook calling it.
Source & flagged code
7 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 unpkgSource creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.
Findings
Affected versions and remediation
This report applies to devez-vibe@1.10.14.
See all versions of devez-vibe for other recorded verdicts.
- Review the evidence and your use of devez-vibe@1.10.14 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.