ClawHub
Back to skill

Security audit

unisound-med-insurance-fee

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed medical-insurance LLM command-line wrapper that sends user-provided questions to a configured model endpoint, so the main risk is handling sensitive claim or patient data carefully.

Install only if you are comfortable sending the provided insurance or medical questions to the configured model service. De-identify real patient, claim, billing, or policyholder data first, use the default or another trusted API endpoint, and avoid untrusted custom system prompts or API URLs.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding
The skill documents capabilities to read local files, write outputs, and send data over the network, but does not declare permissions or clearly bound those actions. In this context, the script can ingest claim documents and forward their contents to a remote LLM endpoint, creating a real risk of unexpected data exposure and policy bypass if operators assume the skill is self-contained and low-privilege.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding
The declared purpose frames the skill as a specialized insurance compliance/calculation tool, but the documented behavior is actually a generic remote LLM wrapper with customizable system prompt, model, and API URL. That mismatch is dangerous because users may trust it with sensitive medical/claims data under the assumption that processing is purpose-limited, while the skill can be repurposed to exfiltrate data or perform unrelated remote inference.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The script transmits user-supplied medical insurance questions, and elsewhere also returns embedded metadata in outputs, to an external HTTPS API using an app key, but provides no consent prompt, redaction step, or disclosure that potentially sensitive healthcare or claim data may leave the local environment. In a medical/insurance context, prompts can contain personal health information, claim identifiers, and billing details, so silent exfiltration to a remote model endpoint materially increases privacy, compliance, and contractual risk.

Static analysis

No suspicious patterns detected.