@vibeapi/api-helper@0.0.139
CLI tool for managing AI coding assistants (Claude Code, Codex, OpenClaw, Hermes, AtomCode, Grok Build) with one unified config
AI Security Review
scanned 8d ago · by lpm-firewall-aiStatic inspection confirms persistent environment changes on ordinary CLI command paths, creating a concrete risk of removing user settings. The evidence supports a warning for destructive capability, but does not establish a malicious attack or credential exfiltration.
Static reason
One or more suspicious static signals were detected.
Trigger
Running a non-configuration CLI command that is not a dry run, after startup update handling.
Impact
Users can lose provider credentials, base URL overrides, or model settings, including values they set for other workflows.
Mechanism
The command calls cleanup code that removes matching Anthropic exports from shell profiles and deletes matching variables from the current process and Windows environment.
AI rationale
Static inspection confirms persistent environment changes on ordinary CLI command paths, creating a concrete risk of removing user settings. The evidence supports a warning for destructive capability, but does not establish a malicious attack or credential exfiltration.
Evidence
dist/lib/command.jsdist/lib/env-cleanup.js
Decision evidence
public snapshotAI called this Suspicious at 95.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- Ordinary non-configuration CLI commands can reach environment cleanup; the cleanup list includes Anthropic API keys and settings.
- The cleanup removes matching exports from shell startup files and deletes matching variables from the running process environment.
- Diagnostic reporting is invoked with the explicit report option; its request uses the configured base URL and API key.
- Claude plugin installation is exposed as a CLI adapter operation, and the installer checks installed plugins and supports an environment opt-out.
Evidence against
- The inspected source establishes persistent environment changes, but no unrelated credential recipient, data theft, or install-time trigger for those changes.
- Telemetry requests use the configured service base URL; the inspected source does not establish that the API key is sent to an unrelated recipient.
Behavioral surface
ChildProcessEnvironmentVarsNetworkShell
Source & flagged code
6 flaggeddist/lib/tools/desktop-process.jsView file
1import{spawnSync as e}from"child_process";import{logger as t}from"../../utils/logger.js";export const WINDOWS_PACKAGE_FAMILIES=["Claude_pzs8sxrjxfjjc","AnthropicPBC.Claude_fnn82j28...
High
Child Process
Package source references child process execution.
dist/lib/tools/desktop-process.jsView on unpkg · L1dist/lib/tools/tool-installs.jsView file
1import{spawnSync as t}from"child_process";import{existsSync as r,realpathSync as n,statSync as e}from"fs";import{delimiter as o,dirname as s,isAbsolute as i,join as a}from"path";im...
High
1import{spawnSync as t}from"child_process";import{existsSync as r,realpathSync as n,statSync as e}from"fs";import{delimiter as o,dirname as s,isAbsolute as i,join as a}from"path";im...
Findings
5 High4 Medium4 Low
HighChild Processdist/lib/tools/desktop-process.js
HighShelldist/lib/tools/tool-installs.js
HighSame File Env Network Executiondist/lib/tools/claude-desktop-cli.js
HighTrigger Reachable Persistencedist/lib/env-cleanup.js
HighRuntime Package Installdist/lib/tools/tool-installs.js
MediumNetwork
MediumEnvironment Vars
MediumInstall Persistencedist/lib/persistent-env.js
MediumStructural Risk Force Deep Review
LowScripts Present
LowHigh Entropy Strings
LowUrl Strings
LowNo License
Affected versions and remediation
This report applies to @vibeapi/api-helper@0.0.139.
See all versions of @vibeapi/api-helper for other recorded verdicts.
- Review the evidence and your use of @vibeapi/api-helper@0.0.139 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.