ClawHub
Back to skill

Security audit

投标机会顾问

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed bidding-analysis assistant that uses public procurement data and an optional local company profile to generate advice and reports.

Before installing, understand that the skill may search or fetch public procurement pages and can store a local company profile at ~/.bidprofile.json if you create one. Use it with bidding information you are comfortable storing locally, review generated reports before sharing them, and avoid providing third-party credentials unless you explicitly intend to use a configured data source.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding
The skill claims it only provides analysis/advice and emphasizes not exfiltrating data or acting without authorization, yet it also instructs the agent to fetch external data, persist a company profile to ~/.bidprofile.json, and generate reports. This mismatch can mislead users and reviewers about actual data handling, causing unauthorized local data persistence or network activity involving sensitive company qualification/profile information under a weaker consent model than users expect.

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The architecture explicitly states the agent will use WebFetch/Bash to retrieve raw HTML and use WebSearch as a fallback, which means the skill can initiate outbound network requests and potentially transmit user queries or company-related context to third-party services. This conflicts with the stated privacy posture of not sending external data without authorization, creating a real risk of unauthorized data disclosure if the implementation does not enforce explicit user consent before every external fetch.

Description-Behavior Mismatch

Medium
Confidence
88% confidence
Finding
The skill metadata says it only performs analysis and does not send data externally without authorization, yet this script contains built-in direct network fetching via `--kw` and detail retrieval. Even if positioned as a fallback path, it enables self-initiated outbound requests that can violate expected trust boundaries, policy constraints, or user consent assumptions in an agent environment.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.