Back to skill

Security audit

护工助手

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent caregiving dashboard, but it handles sensitive patient data with weak local storage and unsafe import/rendering behavior.

Review before installing or using with real patient information. Treat this as a local demo unless privacy controls are added: avoid shared browser profiles, do not import untrusted backup files, protect exported JSON files, and do not rely on it for regulated healthcare records or operational care without stronger access control, validation, audit, and storage protections.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
assets/caregiver.html:412
Finding

Persistent DOM-Based Cross-Site Scripting Through Unsanitized Application Data

Content
View full analysis

👤 患者管理

+ 添加患者
全部责任护工 ${data.team.map(t => `${t.name}`).join('')} 全部护理等级 特级一级二级三级
${patients.map(p => { const caregiver = data.team.find(t => t.id === p.assignedCaregiver); return `
姓名年龄性别床位护理等级责任护工家属联系操作
${p.name}${p.allergies ? '
⚠ '+p.allergies+'' : ''}
${p.age}${p.gender}${p.bedNo || '-'}${p.careLevel}${caregiver ? caregiver.name : '未分配'}${p.familyContact || '-'} ${p.familyPhone || ''}
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
assets/caregiver.html:248
Finding

Sensitive Healthcare and Contact Records Stored Unprotected in Browser Local Storage

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Note
Location
SKILL.md:23
Finding

Skill Requests Unnecessary Shell and Filesystem Modification Capabilities

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (9)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill is explicitly designed to handle highly sensitive information, including patient health records, medication schedules, vital signs, and family emergency contacts, yet it provides no privacy notice, consent guidance, retention policy, or security warning. This is especially dangerous because the skill also states data is stored in browser localStorage, which is not appropriate for sensitive medical data and may be accessible to other local users, scripts, or through insecure backups.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README explicitly states that all data is stored in browser localStorage and supports JSON import/export, but gives no warning that the app handles highly sensitive patient, medication, and health data. In a caregiving context, this can lead operators to store protected health information insecurely on shared devices or export it unencrypted, increasing risk of privacy breaches, unauthorized disclosure, and regulatory noncompliance.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list includes broad terms such as '患者管理', '护理管理', and '生命体征', which can cause the skill to activate in general healthcare conversations that were not intended to invoke this specific tool. In a caregiving context, unintended activation is more risky because it may expose or prompt collection of sensitive patient data or lead users into relying on the wrong workflow.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The export function allows one-click extraction of the entire local dataset, which includes highly sensitive patient, medication, vital-sign, family-contact, and caregiving records. In a healthcare context, exporting PHI/PII without an explicit privacy warning, scope selection, or access control materially increases the risk of accidental disclosure or unauthorized sharing.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The import flow replaces the in-memory dataset with imported content and persists it immediately, without warning the user that existing patient-care records may be overwritten. In this nursing-management context, silent overwrite can lead to loss or corruption of medical, medication, and handover records, which can directly affect continuity of care and patient safety.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The README's user-facing instructions and descriptions are entirely in Chinese, which effectively forces a specific language for users without any visible opt-in or explanation. Under the policy for natural-language constraints, this can be a locale/language restriction unless the skill explicitly states it is intended for a Chinese-speaking or region-specific audience.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

文件标题、描述、触发词和使用说明均仅以中文提供,并通过中文唤起短语定义使用方式,没有提供语言/区域选择或说明该技能仅面向特定语言用户。根据语言/locale 政策,这可能构成未获用户选择的单一语言强制。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The document declares zh-CN and all visible UI text is hard-coded in Simplified Chinese, which imposes a specific language/locale on all users. The file does not provide a language selector, opt-in, or documentation showing that the tool is intentionally limited to a China-specific deployment context.

Content

No source excerpt is available for this finding.