ClawHub
Back to skill

Security audit

智慧招采专家Smart Procurement Expert

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed procurement document assistant that processes user-selected files locally and does not show hidden data access, persistence, or exfiltration behavior.

Install only if you are comfortable giving the skill access to the bid/procurement files you explicitly provide. Review generated drafts, price-score configs, and legal/commercial conclusions manually before relying on them, especially because one parser edge case can fail on unstructured DOCX files.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Lp3

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding
The skill advertises and documents local script-based processing that reads user-supplied procurement files and writes generated DOCX/config outputs, but no permissions are explicitly declared. This creates an authorization and transparency gap: users and the hosting platform cannot easily reason about what filesystem access is needed, increasing the risk of over-broad file access or unsafe handling of sensitive bid materials.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
When no headings are detected, the code sets `cleaned = []` but never initializes `scheme`. The later config-generation path always references `scheme`, which can raise an exception and abort processing on valid-but-unstructured input documents. In this skill context, that is primarily a denial-of-service/reliability flaw rather than a direct security compromise, but it can be triggered by crafted input files to break an automated pipeline.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.