Back to skill

Security audit

Stranger Proximity Alert Skill | 陌生人靠近预警技能

Security checks for vulnerabilities and agentic risk

Overview

This skill may upload surveillance media involving children to an external cloud service and silently create or store account tokens, while its implementation is broader and less clearly scoped than the safety-alert description suggests.

Review this carefully before installing. Use it only with media you are legally allowed to upload, especially where children, bystanders, locations, or school/home footage are visible. Do not rely on it as a complete real-time child-protection system unless the publisher documents the actual alerting path, retention/deletion controls, report access controls, and local token storage behavior.

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
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (58)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The finding describes identity creation/resolution, local file and token persistence, outbound API requests, and authentication handling instead of people/safety-event analysis. In a minor-protection context, this is dangerous because it hides a substantial identity and data-processing subsystem behind a reassuring safety-oriented description, increasing both privacy exposure and misplaced trust.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: "stranger-approach-warning-analysis"
description: "Detects the appearance of strangers near minors and actively issues safety reminder alerts to protect minor safety, suitable for homes, schools, childcare centers, and other scenarios. | 陌生人靠近预警技能,检测未成年人身边出现陌生人员,主动发出安全提醒预警,守护未成年人安全,适用于家庭、学校、托管场所等场景"
version: "1.0.22"
license: "MIT-0"
---

# 🚶 Stranger Proximity Alert Skill | 陌生人靠近预警技能
> **智能分析中枢** · 图片/视频智能分析 · 结构化报告 · 历史报告云端查询

---

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill encourages uploading surveillance images/videos involving minors and querying cloud-stored historical reports without a clear privacy warning, consent model, retention statement, or data-handling disclosure. In this context, that omission is serious because the data is highly sensitive and may include children, bystanders, and location/time information.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The implementation is a generic file/URL submission and report retrieval wrapper, not a dedicated stranger-near-minor warning skill as advertised. This mismatch is dangerous because operators may deploy it expecting real-time child-safety detection and alerting, creating a false sense of protection while sensitive media is instead sent for generic external analysis.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

This file implements a generic API client with arbitrary URL-based GET/POST/PUT/DELETE and CRUD-style helpers, which substantially exceeds the stated purpose of a stranger-near-minors warning skill. In a skill handling sensitive safety data about minors, this overbroad network capability increases the chance of unauthorized data exfiltration, remote state changes, or abuse by other components passing attacker-controlled endpoints or payloads.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The exposed add, edit, delete, http_put, and http_delete methods provide generic remote modification primitives without any constraint tying them to the skill’s narrow detection-and-alert use case. If reachable by other skill logic or untrusted inputs, they enable unauthorized changes or deletion of remote data and expand the attack surface well beyond passive warning analysis.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

The method named delete issues a POST request to perform a deletion action, but the file contains no confirmation, warning, logging, or explanatory comment indicating that it may trigger destructive or irreversible remote changes. Because deletion is safety-critical, the absence of any disclosure is a notable quality and safety gap.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

This file implements generic user-account persistence that is not justified by the advertised 'stranger approaching minors' warning capability. In a safety-focused skill, collecting and managing unrelated account records broadens the attack surface and creates unnecessary retention of personal data, making compromise more damaging and harder to justify.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The User model stores highly sensitive personal and authentication data, including token, open_token, email, birthday, age, and realname, despite no clear need from the skill's stated child-safety alert purpose. Unnecessary storage of identity and token material significantly increases privacy harm and account-compromise impact if the local database is accessed or exfiltrated.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

This utility module contains broad agent workspace discovery, directory creation, local identity-file reading, and default user identity provisioning that are unrelated to a stranger-warning detection skill. In this context, the excess capability expands access to local workspace data and establishes hidden identity state, creating unnecessary attack surface and enabling data access behaviors users would not expect from a child-safety alert skill.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The code silently derives or creates a persistent open-id by reading local files, querying a database, or generating and storing a synthetic user identity. That behavior is unjustified for the described safety-analysis function and can create durable user tracking/account linkage without clear user knowledge or necessity.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The HTTP layer performs authenticated outbound API calls, injects user identifiers into requests, and can trigger implicit account bootstrap/login behavior that is not described by the manifest. For a stranger-approach warning skill, this mismatch is especially concerning because it enables undisclosed external transmission and remote account activity under the guise of a local safety function.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding

The skill advertises and instructs use of shell execution, local file handling, environment access, and outbound network/API calls, but does not declare any explicit tool scope or permissions boundaries. This increases the chance that an agent/runtime grants broader capabilities than users or reviewers expect, making misuse or overreach harder to detect and constrain.

Content

No source excerpt is available for this finding.