The AI review rates the findings as likely false positive, but the risk score (64/100) still counts them.
Analysis record
- Analysed
- 7 months ago 23 April 2026
- Version
- v0.0.12
- Artifact
- SHA256 9c012224…a1dfab
- Detail
- Full list
Evidence ledger
Ranked by severity · findings with a source location link to the code viewer
YARA Rule Matches
9 rules| Severity | Rule | Hits | Files | Metadata |
|---|---|---|---|---|
| HIGH | postinstall file download File download activity detected | 2 | README.mdlib/webdriver-agent.js | Risky Plugins Authors FP 30% |
| HIGH | postinstall obfuscation Code obfuscation techniques detected | 2 | README.mdlib/server.js | Risky Plugins Authors FP 20% |
| HIGH | credential env files Environment configuration file path detected | 3 | lib/logger.jslib/android.jslib/ios.js | Risky Plugins Authors FP 10% |
| HIGH | postinstall network communication Network communication detected | 7 | lib/webdriver-agent.jslib/index.jsREADME.md +4 more | Risky Plugins Authors FP 30% |
| HIGH | postinstall file manipulation File system manipulation detected | 4 | lib/webdriver-agent.jspackage.jsonlib/ios.js +1 more | Risky Plugins Authors FP 20% |
| HIGH | UsingShellInterpreterWhenExecutingOSCommands Arbitrary OS command injection vulnerabilities are more likely when a shell is spawned rather than a new process, indeed shell meta-chars can be used (when parameters are user-controlled for instance) to inject OS commands. For more information checkout the CWE-78 (https://cwe.mitre.org/data/definitions/78.html) advisory. | 1 | lib/iphone-simulator.js | FP 10% |
| HIGH | postinstall system command System command execution detected | 4 | LICENSElib/android.jslib/ios.js +1 more | Risky Plugins Authors FP 10% |
| HIGH | NoUseSocketManually Sockets are vulnerable in multiple ways: They enable a software to interact with the outside world. As this world is full of attackers it is necessary to check that they cannot receive sensitive information or inject dangerous input.The number of sockets is limited and can be exhausted. Which makes the application unresponsive to users who need additional sockets. In many cases there is no need to open a socket yourself. Use instead libraries and existing protocols For more information checkout the CWE-20 (https://cwe.mitre.org/data/definitions/20.html) advisory. | 1 | lib/ios.js | FP 20% |
| HIGH | postinstall crypto operations Cryptographic operations detected | 1 | lib/ios.js | Risky Plugins Authors FP 30% |
Finding Categories
YARA Rules Matched
9 rules(25 hits)MCP Server Analysis
MCP servers expose tools and resources to AI assistants. Unlike browser extensions, they run as standalone processes with direct system access. Tool definitions are analyzed for prompt injection, data exfiltration, and tool poisoning patterns.
AI Security Report
AI Security Review
Evidence context: threat category none; evidence quality moderate.
Tool Poisoning Assessment
The scanner detected zero tool-poisoning findings for @mseep/mobile-mcp. This is the most critical security indicator for MCP packages. Tool poisoning—the defining threat to MCP servers—involves hidden AI manipulation directives embedded in tool descriptions. The absence of any tool-poisoning findings means no hidden instructions like "do not tell the user," "silently execute," or XML-style directive tags were found in tool metadata. The findings_summary explicitly shows "tool-poisoning":"0".
Credential and Network Access
No credential-access findings were detected. The findings_summary shows "secret":"0" with no credential-access findings targeting sensitive paths like .ssh/, .aws/credentials, .kube/config, or environment variables containing secret names (GITHUB_TOKEN, AWS_SECRET_ACCESS_KEY, etc.). This is a strong indicator that the package does not harvest credentials.
Network findings show 10 NET-FETCH detections in lib/webdriver-agent.js at lines 15, 25, 37, 50, 62, 87, 102, 158, 168, and 186. These are generic fetch calls without associated XIOC findings pointing to suspicious domains. The file name webdriver-agent.js aligns with the package's stated purpose as a "Mobile MCP" server—webdriver agents legitimately make HTTP requests to communicate with mobile device emulators or browser automation services. The absence of IoC findings ("ioc":"0" in findings_summary) confirms no suspicious domains or IPs were detected.
Malware Signature Findings
The 25 malware-signature findings (27 high severity total) are YARA code-smell rules firing on normal JavaScript patterns. The distinction between "malware-signature":"25" and "malware":"0" indicates signature matches without confirmed malicious behavior. YARA rules like postinstall_*, credential_*, and code-quality rules commonly trigger false positives on legitimate Node.js code that uses standard libraries (fetch, fs, process.env).
Strongest Counterargument
The 27 high-severity findings could suggest malicious code. However, severity scores in CVEQ are inflated by signature match counts, not behavioral evidence. The absence of tool-poisoning, credential-access, and IoC findings—the three indicators of actual MCP threats—demonstrates these are false positives. The network activity in lib/webdriver-agent.js is consistent with the package's mobile webdriver purpose, not exfiltration architecture.
Conclusion
This package shows no evidence of the defining MCP threats: no tool poisoning, no credential harvesting, no suspicious network destinations. The findings are consistent with YARA code-smell false positives on legitimate mobile automation code.
Key Reasons
- Zero tool-poisoning findings (defining MCP threat is absent)
- Zero credential/secret access findings
- Zero IoC findings (no suspicious domains or IPs)
- Network calls in webdriver-agent.js align with stated mobile MCP purpose
- Malware signatures are YARA code-smell false positives on normal JavaScript
False Positive Considerations
- YARA malware-signature rules firing on legitimate JavaScript code
- Network findings from webdriver-agent.js are expected for mobile MCP functionality
- High severity counts from signature matches rather than behavioral evidence
Reviewed 2026-04-27; recommended action: no action; model confidence 85%.
Source Code Not Available
Source code is not available for this version of the extension.
About This Extension
Frequently Asked Questions
Similar Extensions
Related extensions from the same publisher or marketplace