Security audit

Xiwang Medical Research

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed medical-document search and script-writing helper, but it can recursively read any caller-supplied document directory and uses serious medical claims for promotional output with limited safety guardrails.

Review before installing. Use it only with the intended medical research corpus, avoid pointing it at broad personal or business folders, and have qualified medical/legal reviewers check any disease-treatment or product-promotion scripts before publication.

A.I.G

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

SkillSpector

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (12)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 96% confidence
Finding: The skill description promises a bounded, domain-specific medical research retrieval and script-generation capability, but the documented behavior is broader and inconsistent: it can scan arbitrary passed-in directories and does not actually implement the claimed video-script generation. This mismatch is dangerous because users and reviewers may trust the declared purpose while the operational behavior enables more generic file enumeration/access than advertised.
Content: No source excerpt is available for this finding.

Undeclared Tool Scope

Medium

Category: MCP Least Privilege
Confidence: 87% confidence
Finding: The skill instructs the agent to read from a specific local filesystem path and use a retrieval script, but it does not declare any explicit tool scope or allowed-tools boundary. That creates an authorization ambiguity where the agent may access local files beyond what users or reviewers expect, especially because the referenced directory contains many documents and the script accepts a general directory argument.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The manifest description is entirely in Chinese and the skill instructions throughout the file assume Chinese-language queries, categories, and outputs, but there is no statement that use of Chinese is optional or limited to a region-specific deployment. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 90% confidence
Finding: The index explicitly catalogs a large set of medical-treatment, disease, and therapeutic-claim documents and the skill description says it can generate video scripts and promotional copy from them, yet there is no safety disclaimer, limitation notice, or instruction to avoid diagnosis/treatment claims. In this context, the omission can facilitate generation of misleading or unqualified medical advice or marketing claims, especially around cancer, cardiovascular disease, insomnia, fertility-related conditions, and other sensitive health topics.
Content: No source excerpt is available for this finding.

Ae4

Medium

Category: analysis-evasion
Confidence: 80% confidence
Finding: Suspicious Unicode normalization or mixed-script content
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 97% confidence
Finding: This skill organizes and surfaces disease-treatment claims, efficacy figures, and product application guidance across cancer, gynecology, ENT, sleep, cardiovascular, and TCM contexts without any warning that the content is not medical advice or that claims may be unverified, incomplete, or unsafe for self-treatment. In an agent setting, that makes it easier to generate authoritative-sounding medical recommendations or promotional scripts that could mislead users into delaying care, self-treating serious conditions, or relying on inadequately supported product claims.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: This markdown file contains prescriptive natural-language instructions entirely in Chinese for producing the final script, but it does not indicate that language selection is optional or limited to a China-specific compliance context. Under the policy, forcing a specific language without user opt-in is a natural-language locale violation.
Content: No source excerpt is available for this finding.

Ae4

Medium

Category: analysis-evasion
Confidence: 80% confidence
Finding: Suspicious Unicode normalization or mixed-script content
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The file’s docstring and all user-facing CLI messages are written only in Chinese, which indicates the skill is designed around a fixed language/locale. The policy for this audit flags language constraints when they are imposed without explicit user opt-in or a documented region-specific justification.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The script accepts an arbitrary dirpath from the command line and recursively scans and reads all .docx/.pdf files under it, rather than enforcing the advertised dataset boundary. In an agent setting, this can enable unintended local file discovery and content extraction from other accessible directories if an attacker can influence the path argument.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 97% confidence
Finding: 主入口仅提供 scan、list、search、topic、disease、topics 六类命令，全部围绕文件枚举与内容检索展开，没有任何文案生成、模板渲染或脚本输出逻辑。技能宣称支持根据检索结果自动生成视频口播文案，但该文件实际行为并不覆盖该描述。
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 80% confidence
Finding: The natural-language content and headings are entirely Chinese, and the file does not indicate that this language choice is optional, user-selected, or required for a region-specific purpose. Per SQP-3, forcing a specific language without user opt-in can be a policy violation when no locale justification is provided.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.