Medical Record Translator

• Outcome: Pass
• Latest audit: May 28, 2026 · 6:25 AM UTC
• Version: 1.0.0

Overview

This skill coherently translates user-provided medical records and optionally exports/checks PDFs, with the main risk being expected handling of sensitive health information by the configured AI service.

Install only if Chinese-first medical-record translation is what you want and you are comfortable sending the document contents to the AI provider configured in your agent. De-identify patient information when possible, use an approved local or enterprise model for regulated data, and review the pinned PDF-rendering dependencies before installing them.

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 87% confidence
• Finding: The skill handles user-supplied medical documents and explicitly instructs the system to ingest files, classify PDFs, and extract content, which implies file-read capability without an explicit permission declaration. In a sensitive-medical-data context, undeclared file access weakens transparency and policy enforcement, increasing the risk of unintended access to local or attached files beyond what users expect.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 83% confidence
• Finding: The skill is described as a translation tool, but it also includes export, rendering, and rendered-PDF validation behaviors that are not disclosed in the primary description. This mismatch can hide additional processing of sensitive medical content, causing users or policy systems to underestimate what the skill will do with PHI and making review and consent less reliable.

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The default prompt invokes the skill whenever a user provides a medical record for translation, but it does not constrain language direction, user consent, data sensitivity handling, or explicit confirmation that translation is desired. In a medical-record context, broad invocation increases the chance of processing highly sensitive health information when the user may have intended summarization, explanation, or a different output format, creating privacy and inappropriate-automation risks.

Natural-Language Policy Violations

Medium

• Confidence: 94% confidence
• Finding: The skill description and default prompt hard-code Chinese-first output, which can override user intent and cause disclosure or mishandling of medical information in an unintended language. In medical settings, forced language transformation can create comprehension, safety, and privacy issues if clinicians, patients, or downstream systems expect the original language or a different target language.

Natural-Language Policy Violations

Medium

• Confidence: 92% confidence
• Finding: The contract hard-requires Chinese-first output for all uses, without any indication that this depends on user preference, locale, or downstream policy. In a medical-record translation skill, forcing a fixed language order can cause policy noncompliance, user harm, or unusable outputs when clinicians, patients, or systems require another primary language.

Static analysis

No suspicious patterns detected.