registry / devez-vibe / 1.9.1
devez-vibe@1.9.1
Stable terminal UI for Codex and Claude Agent SDK
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThis is a real install-time agent-extension and environment-mutation risk, so it should be warned rather than marked clean. The inspected implementation is package-aligned and lacks concrete malware behavior, so the first-party extension lifecycle policy supports downgrade rather than blocking.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review; source fingerprint signature matched known malicious package; routed for review; previous stored version diff introduced dangerous source
Trigger
npm postinstall during package installation
Impact
The package adds persistent instructions to AI-agent skill surfaces without an explicit setup command and changes the user's Python package environment.
Mechanism
Automatic agent-skill deployment and user-wide dependency installation
AI rationale
This is a real install-time agent-extension and environment-mutation risk, so it should be warned rather than marked clean. The inspected implementation is package-aligned and lacks concrete malware behavior, so the first-party extension lifecycle policy supports downgrade rather than blocking.
Evidence
package.jsoninstall-skills.mjsprepare-search-deps.mjsskills/luna-loop/SKILL.mdskills/luna-loop/skills/insane-search/$CODEX_HOME/skills/luna-loop$CODEX_HOME/skills/insane-search$CLAUDE_CONFIG_DIR/skills/luna-loop$CLAUDE_CONFIG_DIR/skills/insane-search
Decision evidence
public snapshotAI called this Suspicious at 96.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- The postinstall hook automatically runs two setup scripts.
- The installer copies and prunes bundled skills in both Codex and Claude user skill directories.
- The postinstall setup runs pip with user-wide installation enabled.
- A bundled Codex skill directs the agent to launch a specific delegated model.
Evidence against
- No source path inspected sends credentials or local files to a remote endpoint.
- The bridge explicitly removes Anthropic credential variables before spawning its Claude CLI child.
- The agent-facing files are named package-bundled skills rather than hidden configuration changes.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemShell
HighEntropyStringsUrlStrings
Source & flagged code
10 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node install-skills.mjs && node prepare-search-deps.mjs
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node install-skills.mjs && node prepare-search-deps.mjs
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkginstall-skills.mjsView file
1import {
L2: copyFileSync,
L3: existsSync,
L4: mkdirSync,
L5: readdirSync,
L6: rmSync,
L7: writeFileSync,
L8: } from "node:fs";
...
L16: const userHome = homedir();
L17: const codexHome = process.env.CODEX_HOME?.trim() || join(userHome, ".codex");
L18: const claudeHome = process.env.CLAUDE_CONFIG_DIR?.trim() || join(userHome, ".claude");
L19:
Critical
Ai Agent Control Hijack
Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.
Findings
1 Critical6 High5 Medium4 Low
CriticalAi Agent Control Hijackinstall-skills.mjs
HighInstall Time Lifecycle Scriptspackage.json
HighPayload In Excluded Dirskills/insane-search/tests/live_benchmark/bench.py
HighKnown Malware Source Similaritybridge/claude-agent-sdk-bridge.mjs
HighKnown Malware Source Similaritybin/dvz.js
HighKnown Malware Source Fingerprint Signaturebridge/claude-agent-sdk-bridge.mjs
HighPrevious Version Dangerous Deltabridge/claude-agent-sdk-bridge.mjs
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumEnvironment Vars
MediumShips Native Binarybin/dvz.exe
MediumShips Build Helperskills/insane-search/tests/live_benchmark/bench.py
Affected versions and remediation
This report applies to devez-vibe@1.9.1.
See all versions of devez-vibe for other recorded verdicts.
- Review the evidence and your use of devez-vibe@1.9.1 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.