ClawHub
Back to skill

Security audit

yanmu-dcf-model

Security checks across malware telemetry and agentic risk

Overview

This DCF modeling skill is coherent and disclosed, with expected live quote fetching but no hidden credential use, persistence, or destructive behavior.

Install only if you are comfortable with the skill querying Sina Finance for ticker prices during runs and installing the documented Python dependencies. Use a controlled output directory if you do not want heatmap images written into the current folder.

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 (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding
The skill documentation indicates live price retrieval from Sina Finance and explicitly describes automatic runtime network access, but no corresponding permission declaration is present. Undeclared network capability is dangerous because it expands the skill's trust boundary, can transmit user inputs externally, and prevents users or the platform from making an informed consent decision about outbound data access.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The declared purpose centers on DCF modeling, but the skill also performs external HTTP requests for live market data, which is a material behavioral expansion beyond pure local financial modeling. This mismatch is risky because users may invoke the skill assuming only local computation, while the skill actually contacts third-party services, creating privacy, integrity, and supply-chain exposure through unannounced external dependencies.

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The script performs an undeclared outbound network request to Sina Finance to override locally stored prices. In a skill expected to behave as an internal/offline DCF model, this creates a trust-boundary violation: results depend on external data availability and integrity, and execution may leak usage metadata or fail unpredictably in restricted environments.

Intent-Code Divergence

Medium
Confidence
88% confidence
Finding
The code comments describe the data as internal/shared company data, but the script also fetches live market data externally. This mismatch can mislead reviewers and operators about data provenance and network behavior, increasing the risk of deploying the skill in environments where outbound access or untrusted data sources are unacceptable.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.