ClawHub
Back to skill

Security audit

A股个股深度分析

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed A-share stock analysis helper that fetches public market data, with some data-quality and privacy caveats but no evidence of malicious behavior.

Install only if you are comfortable with the agent querying public Chinese finance sites and web search providers for the stocks you ask about. Treat generated buy/sell points and target prices as informational analysis, not investment advice, and check that missing financial-statement data is marked unavailable rather than inferred.

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 (2)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The skill advertises use of an internal a-stock-data package but actually instructs the agent to contact multiple third-party endpoints and to fall back to generic web_search/web_crawl. That expands the trust boundary, can leak user query context or stock-interest metadata to external services, and bypasses the tighter controls implied by a packaged data source.

Description-Behavior Mismatch

Medium
Confidence
77% confidence
Finding
The skill description and report template promise analyses such as order visibility, cash-flow matching, capacity expansion drivers, and detailed financial-statement-based judgments, but the embedded collection script does not gather the underlying data needed for those conclusions. This creates a risk that the agent will fabricate or over-infer unsupported financial analysis, undermining integrity and potentially misleading users in a high-stakes investing context.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.