ClawHub
Back to skill

Security audit

Protein Docking Configurator

Security checks across malware telemetry and agentic risk

Overview

This is a local molecular-docking helper that reads user-provided structure files and writes docking config files, with no evidence of hidden network, credential, persistence, or destructive behavior.

Install only if you are comfortable running a local Python script on protein or ligand files. Run it in a project workspace, provide explicit input and output paths, and avoid pointing the output at existing files you do not want overwritten.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Lp3

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding
The skill documentation describes file read and file write behavior but does not declare any corresponding permissions or enforceable constraints. This creates a mismatch between stated capabilities and the trust boundary, which can lead to unauthorized access to arbitrary local paths if the implementation does not strictly validate input and output locations.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.