Back to skill

Security audit

AI大模型专家|TVC 广告片

Security checks for vulnerabilities and agentic risk

Overview

This skill mostly matches its video-generation purpose, but it can send API credentials and media to external endpoints and has insufficient safeguards around automatic invocation, custom API hosts, and downloaded result paths.

Install only if you are comfortable with AI-HIVE receiving the prompts and media files you provide, and avoid enabling automatic use for vague requests. Keep the API base URL at the documented AI-HIVE endpoint unless you deliberately trust another endpoint, protect the API key, review generated commands before running them, and choose an output directory where accidental overwrites would not matter.

Vulnerability Patterns
  • 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
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/videogen.py:91
Finding

Configurable API Base URL Can Expose the Bearer Credential to an Arbitrary Server

Content
View full analysis
` header and attaches it to every request made to the selected base URL. Consequently, an arbitrary server selected through any of these configuration channels receives the AI-HIVE credential as well as request content such as prompts, model selections, pricing data, and task identifiers. Allowing custom endpoints can be legitimate for development or self-hosted deployments, but sending the same production credential to an un ...[truncated 1579 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/videogen.py:399
Finding

Untrusted Task Metadata Is Used to Construct Download Paths

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
Findings (10)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding

The skill documents and appears to invoke sensitive capabilities including shell execution, network access, environment/config access, and file reads/writes, but it does not declare any explicit tool scope or permission boundaries. That creates an overprivileged integration surface where a host agent may expose more capability than users expect, increasing the risk of unintended file access, secret handling, network exfiltration, or command execution through normal skill use.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest sets metadata.language: zh-CN, and the document content is primarily Chinese, but there is no indication that users may opt into another language or that the skill is limited to a China-specific compliance or regional use case. This creates a natural-language policy concern because it appears to enforce a locale without explicit user choice.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill enables implicit invocation without any documented trigger constraints, exclusions, or user-confirmation guardrails. In a skill that can initiate real AI-HIVE workflows involving uploads, task submission, polling, and downloads, this increases the chance of the agent invoking external operations unexpectedly based on vague user intent, which can lead to unintended data transfer, cost-incurring actions, or workflow execution.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest describes a skill for TVC/video generation workflows through AI-HIVE, but this file’s documented and implemented functionality also exposes unrelated generic AI-Hive capabilities such as text chat, image generation, model enumeration, and wallet/user-info queries. Those capabilities are broader than the stated purpose of a TVC video skill and indicate the code is a repurposed general AI client rather than a narrowly scoped video skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
80% confidence
Finding

Natural-language strings throughout the file, including the top-level description and CLI guidance, are exclusively in Chinese, which effectively forces a specific language for user interaction. The file does not offer an opt-in language choice or explain that the skill is intentionally limited to a Chinese-speaking context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill uploads user-supplied image, video, and audio files to remote AI Hive/object-storage endpoints, but the code provides no explicit consent, privacy notice, destination summary, or confirmation before transfer. In a media-production context, inputs commonly contain unreleased campaign assets, personal likenesses, voice data, or proprietary creative materials, so silent transmission increases the risk of unintentional data disclosure.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The display name, description, and default prompt are written entirely in Chinese, which effectively fixes the skill's language/locale behavior without indicating user choice or a documented regional constraint. This may violate language/locale policy where user opt-in or explicit justification is required.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The example API key placeholder uses Chinese text in the string literal, which imposes a specific language presentation without any opt-in or alternative locale. This can violate the language/locale policy for generally distributed skills when no user choice or justification is provided.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The top-level docstring describes the file as a '通用 AI 能力调用工具' that wraps text chat, image generation, video generation, model query, media upload, and task polling. However, the packaged skill metadata at the bottom defines a TVC-video-specific skill with constrained video-oriented modes, so the inline documentation materially describes a broader intent than the actual skill packaging.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The manifest emphasizes generating TVC/video content, with upload/poll/download as implementation details of that workflow. This wrapper also exposes standalone task and upload commands as separate user-invocable capabilities, which expands the behavior from a focused creation skill into a more generic asset/task management tool.

Content

No source excerpt is available for this finding.