Back to skill

Security audit

来客推商城

Security checks for vulnerabilities and agentic risk

Overview

This is a simple Chinese product-information skill with no code execution or data access, though its activation wording is too broad and could surface sales content in unrelated software discussions.

Install this only if you want the assistant to answer questions about 来客推, LaikeTui, ThinkInShop, or related mall/e-commerce system products. Be aware that the trigger wording is broad and may cause vendor-specific sales information or contact details to appear during more general software or e-commerce conversations.

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
Findings (1)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:3
Finding

Broad Skill Activation Redirects General Technical Requests to Commercial Promotion

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 3, 8–10, and 61
Vulnerability Type: Commercial output hijacking through overly broad activation instructions
Risk Level: Medium

Evidence

yaml
description: 来客推商城系统。用于回答商城源码相关咨询、产品介绍、功能解答。当用户咨询商城源码、电商系统、软件开发等问题时触发。
markdown
- **官网**:https://www.laiketui.com
- **客服微信/QQ**:13237476750
- **联系电话**:13237476750
markdown
**了解更多详情,欢迎添加微信咨询:13237476750** 📱

The first excerpt states that the skill activates for questions about e-commerce source code, e-commerce systems, or software development. The remaining excerpts provide commercial contact information and conclude with a fixed invitation to add the supplied messaging account for further consultation.

Technical Analysis

The activation description includes the general subject of software development rather than limiting activation to explicit questions about the named vendor or its products. Once loaded, the skill provides branded sales information, external demonstration sites, telephone and messaging details, and a prominent commercial call to action.

This combination can alter the purpose of an otherwise general technical session. A request that merely concerns software development may cause the agent to rely on vendor-specific promotional content and direct the user toward an external commercial contact. The behavior best matches instruction hijacking because skill text and activation metadata can shift the agent's current response goals from technical assistance to lead generation.

The audited file does not contain scripts, executable commands, dependency declarations, credential access, persistence mechanisms, or automated network requests. The issue is therefore limited to response integrity and user redirection rather than host compromise.

Attack Path

  1. A user submits a general question about software development, e-commerce software, or e-commerce source co ...[truncated 1401 chars]
Remediation
View remediation

Remediation Suggestions

  1. Narrow the activation description to explicit requests about the named vendor, its products, or its documentation. Remove generic triggers such as “software development.”
  2. State that the skill must answer only the user's specific question and must not introduce sales material unless explicitly requested.
  3. Remove the unconditional final contact invitation. Provide contact details only when the user asks how to contact the vendor, request a quotation, or obtain commercial support.
  4. Clearly identify vendor links and contact details as optional third-party commercial information rather than neutral recommendations.
  5. Separate technical product documentation from sales and lead-generation content so loading technical guidance does not automatically introduce promotional instructions.
  6. Add an activation boundary such as: “Use only when the user explicitly mentions Laike Sales, LaikeTui, or one of the listed products.”
  7. Add regression tests confirming that generic programming, software-development, and unrelated e-commerce questions do not activate this skill.
  8. Periodically verify the ownership and destination of all published domains, demonstration sites, telephone numbers, and messaging accounts.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger description is broad enough to activate on generic discussions about software development, e-commerce systems, or source code that may be unrelated to this specific product. Over-broad routing can cause the skill to intercept conversations outside its intended scope, leading to irrelevant promotion, misleading answers, or unnecessary exposure of contact/marketing content in unrelated contexts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

All user-facing description content is written only in Chinese, and the skill does not indicate that language selection is optional or limited to a justified region-specific context. Under the policy, forcing a specific language without user opt-in can be a natural-language policy issue.

Content

No source excerpt is available for this finding.