Loading npm security reports…

devez-vibe@1.6.46: AI-agent control-surface risk | LPM Firewall

registry / devez-vibe / 1.6.46

devez-vibe@1.6.46

Stable terminal UI for Codex and Claude Agent SDK

AI Security Review

scanned 22d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. npm install runs a postinstall script that writes the bundled luna-loop skill into Codex and Claude user skill directories and removes extra files already in those folders. Those paths are foreign agent control surfaces, not a first-party extension directory owned by this package.

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

Trigger

npm install of devez-vibe, including a global install, which runs the postinstall script.

Impact

Codex and Claude Code can discover and load a third-party skill planted in the user's agent skill paths without a separate setup command.

Mechanism

unconsented install-time copy into Codex and Claude skill directories

Policy narrative

Installing devez-vibe runs postinstall, which copies skills/luna-loop into the user's Codex and Claude skill folders and prunes files that are not in the package copy. That mutates other products' agent control surfaces during npm install. The TUI binary and Claude SDK bridge are separate user-invoked runtime code and are not required for this install-time write.

AI rationale

The postinstall script writes into ~/.codex/skills and ~/.claude/skills, which is unconsented install-time mutation of foreign AI-agent control surfaces. That matches a publish block even though the planted skill is not an exfiltration prompt.

Evidence

package.jsoninstall-skills.mjsskills/luna-loop/~/.codex/skills/luna-loop~/.claude/skills/luna-loop

Decision evidence

Source & flagged code

7 flagged · loading source

package.jsonView file

•scripts.postinstall = node install-skills.mjs

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

40"scripts": { "postinstall": "node install-skills.mjs" },

High

Suspicious Lifecycle Evidence

postinstall automatically runs node install-skills.mjs on npm install.

package.jsonView on unpkg · L40

install-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

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

install-skills.mjsView on unpkg · L1

14const userHome = homedir(); const codexHome = process.env.CODEX_HOME?.trim() || join(userHome, ".codex"); const claudeHome = process.env.CLAUDE_CONFIG_DIR?.trim() || join(userHome, ".claude"); const targets = [ { name: "Codex", path: join(codexHome, "skills", "luna-loop") }, { name: "Claude", path: join(claudeHome, "skills", "luna-loop") }, ];

High

Ai Review Evidence

The install script copies the bundled luna-loop skill into the Codex user skill directory (~/.codex/skills/luna-loop, or CODEX_HOME).

install-skills.mjsView on unpkg · L14

bin/dvz.exeView file

•path = bin/dvz.exe kind = native_binary sizeBytes = 24402432 magicHex = [redacted]

Medium

Ships Native Binary

Package ships native binary artifacts.

bin/dvz.exeView on unpkg

bridge/claude-agent-sdk-bridge.mjsView file

•matchType = normalized_sha256 matchedPackage = devez-vibe@1.6.48 matchedPath = bridge/claude-agent-sdk-bridge.mjs matchedIdentity = npm:ZGV2ZXotdmliZQ:1.6.48 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High

Known Malware Source Similarity

Source file is highly similar to a previously finalized malicious package; route for source-aware review.

bridge/claude-agent-sdk-bridge.mjsView on unpkg

•matchType = malicious_source_fingerprint_signature signature = 1466de81dd2a8cd0 signatureType = suspicious_hashes sourceLabel = final_verdict:malicious matchedPackage = devez-vibe@1.6.48 matchedPath = bridge/claude-agent-sdk-bridge.mjs matchedIdentity = npm:ZGV2ZXotdmliZQ:1.6.48 similarity = 1.000 shingleOverlap = 2 summary = package final verdict is malicious

High

Known Malware Source Fingerprint Signature

Source fingerprint signature matches a known malicious package signature; route for source-aware review.

bridge/claude-agent-sdk-bridge.mjsView on unpkg

Source & flagged code

7 flagged

package.jsonView file

•scripts.postinstall = node install-skills.mjs

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

40"scripts": { "postinstall": "node install-skills.mjs" },

High

Suspicious Lifecycle Evidence

postinstall automatically runs node install-skills.mjs on npm install.

package.jsonView on unpkg · L40

install-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

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

install-skills.mjsView on unpkg · L1

14const userHome = homedir(); const codexHome = process.env.CODEX_HOME?.trim() || join(userHome, ".codex"); const claudeHome = process.env.CLAUDE_CONFIG_DIR?.trim() || join(userHome, ".claude"); const targets = [ { name: "Codex", path: join(codexHome, "skills", "luna-loop") }, { name: "Claude", path: join(claudeHome, "skills", "luna-loop") }, ];

High

Ai Review Evidence

The install script copies the bundled luna-loop skill into the Codex user skill directory (~/.codex/skills/luna-loop, or CODEX_HOME).

install-skills.mjsView on unpkg · L14

bin/dvz.exeView file

•path = bin/dvz.exe kind = native_binary sizeBytes = 24402432 magicHex = [redacted]

Medium

Ships Native Binary

Package ships native binary artifacts.

bin/dvz.exeView on unpkg

bridge/claude-agent-sdk-bridge.mjsView file

•matchType = normalized_sha256 matchedPackage = devez-vibe@1.6.48 matchedPath = bridge/claude-agent-sdk-bridge.mjs matchedIdentity = npm:ZGV2ZXotdmliZQ:1.6.48 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High

Known Malware Source Similarity

Source file is highly similar to a previously finalized malicious package; route for source-aware review.

bridge/claude-agent-sdk-bridge.mjsView on unpkg

•matchType = malicious_source_fingerprint_signature signature = 1466de81dd2a8cd0 signatureType = suspicious_hashes sourceLabel = final_verdict:malicious matchedPackage = devez-vibe@1.6.48 matchedPath = bridge/claude-agent-sdk-bridge.mjs matchedIdentity = npm:ZGV2ZXotdmliZQ:1.6.48 similarity = 1.000 shingleOverlap = 2 summary = package final verdict is malicious

High

Known Malware Source Fingerprint Signature

Source fingerprint signature matches a known malicious package signature; route for source-aware review.

bridge/claude-agent-sdk-bridge.mjsView on unpkg