registry / devez-vibe / 1.7.45
devez-vibe@1.7.45
Stable terminal UI for Codex and Claude Agent SDK
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThe package is not proven to steal data or execute a remote payload, but its install-time global agent mutation and automatic permission-bypass behavior are unsafe. This warrants a warning rather than a publish block under the first-party agent-extension policy.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
npm postinstall, followed by use of the installed skill or Claude bridge.
Impact
It broadens AI-agent control surfaces without a separate setup command and can reduce normal permission safeguards.
Mechanism
Automatic global agent-skill deployment and permission-bypass session setup.
AI rationale
The package is not proven to steal data or execute a remote payload, but its install-time global agent mutation and automatic permission-bypass behavior are unsafe. This warrants a warning rather than a publish block under the first-party agent-extension policy.
Evidence
package.jsoninstall-skills.mjsbridge/claude-agent-sdk-bridge.mjsskills/luna-loop/SKILL.md.codex/skills/luna-loop.claude/skills/luna-loop.claude/settings.json
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- Installation automatically runs a script that deploys an agent skill.
- The installer writes into both global Codex and Claude skill directories and removes files absent from its packaged skill.
- The Claude bridge automatically attempts bypassPermissions for each session.
- The deployed skill instructs the host to run Codex CLI subprocesses.
Evidence against
- No network endpoint or credential exfiltration was found in inspected source.
- The bridge removes Anthropic credential variables from its child environment.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemShell
HighEntropyStrings
Source & flagged code
7 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node install-skills.mjs
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg40"scripts": {
"postinstall": "node install-skills.mjs"
}
Medium
Ai Review Evidence
Installation automatically runs a script that deploys an agent skill.
package.jsonView on unpkg · L40install-skills.mjsView file
1import {
L2: copyFileSync,
L3: existsSync,
L4: mkdirSync,
L5: readdirSync,
...
L14: const userHome = homedir();
L15: const codexHome = process.env.CODEX_HOME?.trim() || join(userHome, ".codex");
L16: const claudeHome = process.env.CLAUDE_CONFIG_DIR?.trim() || join(userHome, ".claude");
L17:
...
L23: function copyTree(source, destination) {
L24: mkdirSync(destination, { recursive: true });
L25: for (const entry of readdirSync(source, { withFileTypes: true })) {
Critical
Ai Agent Control Hijack
Findings
1 Critical2 High6 Medium3 Low
CriticalAi Agent Control Hijackinstall-skills.mjs
HighInstall Time Lifecycle Scriptspackage.json
HighPrevious Version Dangerous Deltabridge/claude-agent-sdk-bridge.mjs
MediumEnvironment Vars
MediumShips Native Binarybin/dvz.exe
MediumStructural Risk Force Deep Review
MediumAi Review Evidencepackage.json
MediumAi Review Evidenceinstall-skills.mjs
MediumAi Review Evidenceinstall-skills.mjs
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
Affected versions and remediation
This report applies to devez-vibe@1.7.45.
See all versions of devez-vibe for other recorded verdicts.
- Review the evidence and your use of devez-vibe@1.7.45 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.