ClawHub
Back to skill

Security audit

Coffee Prices by City

Security checks across malware telemetry and agentic risk

Overview

This skill is a small local coffee price estimator with a disclosed optional IP-based city lookup, not a hidden or destructive tool.

Install only if rough reference prices are acceptable. Provide --city explicitly, or set OPENCLAW_CITY, if you do not want the script contacting ipinfo.io. For sensitive environments, pin the requests dependency before installation and treat the generated prices as estimates rather than authoritative pricing.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

High
Confidence
98% confidence
Finding
The skill advertises fetching current coffee prices from major chains, but the implementation only returns synthetic estimates derived from hardcoded base prices and simple city-tier multipliers. This is a deceptive capability mismatch that can mislead users or downstream agents into treating fabricated data as current real-world pricing, undermining integrity and trust.

Context-Inappropriate Capability

Medium
Confidence
93% confidence
Finding
The code sends a request to ipinfo.io to infer the user's city when none is supplied, disclosing the host's IP address and related metadata to a third party. For a coffee-price comparison skill, this is not necessary to fulfill the stated purpose and creates avoidable privacy and network-exposure risk, especially in enterprise or agent-runtime environments.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
At the point where city resolution falls back to IP-based inference, the user is not clearly informed that an outbound request to a third-party geolocation service may occur. This weakens informed consent and can surprise operators in restricted environments where unannounced external calls are sensitive.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.