Back to skill

Security audit

月嫂助手

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent local maternal-care dashboard, but it handles sensitive mother and infant health data with weak privacy disclosure and unsafe browser-side rendering/import behavior.

Review before installing. Do not use this on shared devices or for regulated medical records unless the publisher adds clear privacy notices, safer export handling, input/output escaping, import validation, and removes unnecessary Bash access. Avoid importing JSON backups from untrusted sources.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
assets/nanny.html:445
Finding

Persistent DOM-Based Cross-Site Scripting Through Unescaped Care Records

Content
View full analysis
{ h += ''+r.time+''+r.type+''+r.amount+''+(r.duration||'--')+''+(r.burped?'✅ 已拍':'--')+''+(r.notes||'')+'删除'; h += ''; }); $('#main-content').innerHTML = h; ``` The same unsafe rendering pattern is used for diaper notes, cord-care notes, bathing notes, maternal-care notes, meal descriptions, medication information, diary entries, and profile values. ### Technical Analysis The application treats stored and imported data as trusted HTML. Fields such as `notes`, `food`, record types, medication data, names, and identifiers can originate from form input or an arbitrary JSON backup. These values are inserted into HTML strings without output encoding and are then parsed by the browser through `innerHTML`. ...[truncated 2275 chars]
Remediation
View remediation

T03 · Remote Payload Retrieval and Execution

Warning
Location
assets/nanny.html:7
Finding

Unverified Remote JavaScript Dependency Executes in the Application Origin

Content
View full analysis
``` ### Technical Analysis The application retrieves and executes JavaScript from a third-party CDN whenever the page is loaded with network access. Although the dependency version is specified, the script element does not use Subresource Integrity. The browser consequently has no locally declared cryptographic digest against which to verify the returned file. If the CDN, its delivery path, the hosted artifact, or an upstream publishing account is compromised, altered JavaScript can execute with the same browser-origin privileges as the application. This is particularly significant because the application stores maternal and newborn identity and health information in `localStorage`. No malicious behavior was found in the project’s own copy of the HTML, and the audit did not establish that the currently served Chart.js resource is malicious. The issue is the avoidable remote trust and supply-chain exposure. ### Attack Path 1. An attacker compromises the CDN delivery path, hosted dependency artifact, or an account capable of replacing the referenced resource. 2. The user opens the care dashboard while connected to the network. 3. The browser retrieves the modified JavaScript because the page explicitly references the remote CDN. 4. Since no `integrity` digest is present, the browser does not detect that the returned content differs from the reviewed artifact. 5. The modified script executes in the page context. 6. It can access application state and `localStorage`, alter records, or transmit sensitive information using browser network APIs. ### Impact Assessment A successful supply-chain attack would execute arbitrary JavaScript in the application's brows ...[truncated 532 chars]
Remediation
View remediation
``` The digest must be obtained from a trusted source or calculated from an independently verified local artifact. 3. Use an explicit Content Security Policy that limits scripts to approved local files or narrowly allowlisted origins and disallows unexpected script execution. 4. Pin and periodically review dependency versions, provenance, known vulnerabilities, and cryptographic hashes. 5. Ensure the application fails safely if the dependency cannot be verified or loaded. Core record access should not require execution of unverified remote code. ]]>
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (10)

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.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest lists generic triggers such as "母婴护理", "新生儿护理", "月子餐", and "产后恢复", which are broad topic phrases rather than narrowly scoped invocation commands. These common domain terms could cause unintended activation whenever a user discusses routine postpartum or newborn care, especially because no exclusion conditions or negative examples are provided.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill repeatedly describes itself as a pure front-end, localStorage-based maternal-care app, yet it requests Bash access. This capability mismatch violates least privilege and creates unnecessary command-execution exposure: if invoked by the agent or later expanded, Bash could access local files, execute arbitrary shell commands, or exfiltrate sensitive maternal and infant health data beyond the stated design.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Bash execution is not justified by the stated purpose of tracking care records, generating reports, and rendering a local HTML dashboard. In a skill handling sensitive postpartum and newborn data, unnecessary shell access materially increases risk because it enables actions outside the user's expected trust boundary, including file system reads, process execution, and bulk data export.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill collects and stores sensitive maternal and infant health information, including medication, body metrics, and care logs, but does not present a clear privacy warning, retention notice, or security guidance. In this domain, missing privacy transparency can lead users to store regulated or highly sensitive health data without understanding local exposure, backup/export risks, or who may access the device/browser storage.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill stores maternal and infant health data, medication records, and care logs in localStorage without any visible notice, consent, retention guidance, or protection. On shared devices or compromised browsers, this sensitive data can be trivially accessed by anyone with local browser access, increasing privacy and confidentiality risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The export function downloads the full dataset as JSON, including sensitive maternal and infant care records, without a warning, confirmation of sensitivity, or scoped export options. Once exported, the data can be copied, forwarded, synced, or left on disk unprotected, which meaningfully increases leakage risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
72% confidence
Finding

The skill's name, description, triggers, and user instructions are entirely in Chinese except for one English trigger, and there is no statement that the skill is China-specific or that users may choose another language. Under the stated policy, forcing a specific language without opt-in can be a natural-language policy issue unless the locale constraint is clearly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The page declares lang="zh-CN", and all user-facing content is written only in Simplified Chinese. Under the stated policy, forcing a specific language/locale without user opt-in is a natural-language policy issue unless the locale limitation is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The manifest describes a postpartum/nanny workflow assistant focused on care management, logging, monitoring, and report generation. While generating family reports is in scope, directly writing report contents to the system clipboard is an extra client-side data-transfer capability that is not explicitly part of the stated purpose and can expose sensitive mother/baby health data beyond the app UI.

Content

No source excerpt is available for this finding.