Back to skill

Security audit

AI医学报告解读助手

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent medical-report helper, but it handles sensitive health records and can publish health-derived content without enough privacy, consent, or redaction safeguards.

Review this skill carefully before installing. Use it only with reports you are comfortable processing through AI and search tools, remove names, IDs, hospital numbers, dates, and other identifiers where possible, and do not use the publication features unless you have manually checked that no private medical information remains.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill handles medical reports, extracts their contents, and queries external services, but it does not clearly disclose privacy risks, data flows, retention, or consent requirements for highly sensitive health information. This is dangerous because users may upload personally identifiable medical records without understanding that their data could be sent to third-party tools or used in downstream publication features.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The workflow processes highly sensitive medical-report data and sends derived abnormal-indicator information to external services such as search and possibly downstream content tools, but it provides no privacy notice, consent boundary, or data-minimization controls. In a medical context this is especially dangerous because health data is highly sensitive, and users may not realize their report contents could leave the local analysis path and be exposed to third-party systems.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The optional publication step allows health reports and knowledge cards to be posted publicly without an explicit warning that the content may reveal personal medical information. Because the workflow concerns health interpretation, even a summarized report or card can disclose conditions, abnormal test results, or other sensitive inferences, creating a significant privacy and reputational risk if published unintentionally.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly supports publishing health-report-derived content to public platforms like WeChat and Xiaohongshu, but the description does not include a clear privacy warning, consent check, or de-identification requirement before sharing. Because the underlying input is a medical report containing highly sensitive personal health information, users may inadvertently expose protected data or enough context to re-identify themselves when generating or posting summaries and cards.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The trigger phrases are very broad health-related terms such as '医学分析' and '健康分析', which can cause the skill to activate for a wide range of general medical conversations beyond the user's intent. In a medical context, accidental activation is riskier because it may solicit or process sensitive health data and provide quasi-medical guidance when the user did not explicitly request this workflow.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The workflow expands from medical-report interpretation into optional public content distribution via WeChat/Xiaohongshu, which is outside the core medical-analysis purpose and materially increases data-exposure risk. In a healthcare context, even derived summaries and knowledge cards may contain personally sensitive health information, making this feature dangerous if enabled without strict separation, consent, and sanitization.

Content

No source excerpt is available for this finding.