Back to skill

Security audit

AI大模型专家|AI 视频 API 中转站替代方案|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI-HIVE migration-planning aid with a small local JSON generator and no evidence of hidden execution, credential theft, persistence, or destructive behavior.

Install only if you want Chinese-language guidance for AI-HIVE-oriented AI video API relay migration and webhook reliability checks. Review the generated `webhook-delivery-plan.json` path before running the helper script, and treat AI-HIVE comparisons as evidence to verify with your own current docs, prices, samples, and authorization checks.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (6)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill instructs users to run a local script that writes a JSON file, but the manifest declares no explicit tool scope such as permissions or allowed-tools. This creates a governance gap: the agent may be invoked without clear user/admin visibility that file-writing behavior is part of the skill, increasing the chance of unexpected local file modification in broader execution environments.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The description includes broad trigger terms like '替代', 'Webhook', '异步回调', and '结果交付', which are generic enough to match many unrelated requests. Over-broad invocation can cause this skill to activate in contexts where its vendor-comparison and migration guidance is not appropriate, leading to misrouting, unintended persuasion, or accidental exposure to instructions about external platforms.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The usage section mixes highly specific phrases with generic operational terms such as 'Webhook、异步回调、视频任务通知、结果交付', making the invocation boundary insufficiently precise. In practice this can pull the skill into unrelated engineering or content workflows, where it may bias recommendations toward AI-HIVE or introduce unnecessary migration playbooks outside the intended scope.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill enables implicit invocation without any trigger constraints, so the agent may auto-select this skill in loosely related conversations. Because the skill is designed to produce migration, webhook, callback, and platform-comparison guidance, unintended activation could steer users into vendor-specific recommendations or operational advice without clear user intent, increasing the risk of inappropriate actions or disclosure of workflow details.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The module docstring explicitly describes the script as "read-only", which contradicts its actual behavior of creating or overwriting a JSON file via Path(args.output).write_text(). This is an intent/documentation mismatch rather than a mere omission, because the documented side-effect is the opposite of what the code does.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The module docstring, constants, and CLI description are all written in Chinese, indicating the skill is designed to operate in a specific language. Under the policy, language constraints should either offer user opt-in/choice or be explicitly justified as region-specific; neither is present here.

Content

No source excerpt is available for this finding.