Security audit

AI电商选品上货 | 青虎AI

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed e-commerce data and listing workflow that uses Qinghu APIs with user authorization, though users should understand it can spend credits, use an API token, export files, and perform listing-related actions when approved.

Install this only if you intend to use Qinghu's e-commerce data workflows. Before approving calls, check the listed tools, expected credit cost, platform/account involved, and whether results or exports may contain sensitive business research. For listing workflows, confirm the exact store/account, links, template, and intended changes before approving.

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

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 (3)

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The top-level description says the skill 'must trigger' for a very broad set of common e-commerce requests across many platforms, with no clear exclusions or priority rules. This can cause inappropriate routing and unintended tool use for loosely related user requests, increasing the chance of unnecessary external data access, billing, and mishandling of user intent.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The '何时触发' examples are broad enough to overlap with ordinary conversational assistance, especially generic requests like '帮我选品' or '这个类目能做吗'. Without stronger scope boundaries, the agent may invoke this skill when the user did not intend external tool calls or paid data access, creating consent and overreach risks.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill instructs automatic export of datasets and reuse of cached local file paths without requiring a clear upfront warning that user-related data will be written to disk. This creates a data handling risk because potentially sensitive business queries, product research, or account-linked results may be persisted locally without informed consent or retention controls.

Static analysis

No suspicious patterns detected.