Back to skill

Security audit

数字人试穿展示视频|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its virtual try-on video purpose, but it can send a user's AI-HIVE API key to an unrestricted custom API endpoint.

Review before installing. Use only the default AI-HIVE endpoint, avoid setting --base-url or AI_HIVE_BASE_URL unless you fully trust the destination, and consider a spend-limited or revocable API key. Confirm all billable generation parameters before running commands, and install dependencies in an isolated environment with pinned packages where possible.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/videogen.py:91
Finding

Bearer API credential can be disclosed through an unrestricted custom Base URL

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
scripts/videogen.py:16
Finding

Runtime dependency installation guidance lacks version and integrity pinning

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (16)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

This finding indicates the skill may expose broader undeclared functionality such as general chat, image generation, account/balance retrieval, model enumeration, or interactive API-key setup while presenting itself as a narrow try-on workflow. Undeclared capabilities are risky because they expand the attack surface, may encourage secret handling or account probing, and defeat policy decisions based on the declared narrow purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

This finding indicates the skill may expose broader undeclared functionality such as general chat, image generation, account/balance retrieval, model enumeration, or interactive API-key setup while presenting itself as a narrow try-on workflow. Undeclared capabilities are risky because they expand the attack surface, may encourage secret handling or account probing, and defeat policy decisions based on the declared narrow purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

This finding indicates the skill may expose broader undeclared functionality such as general chat, image generation, account/balance retrieval, model enumeration, or interactive API-key setup while presenting itself as a narrow try-on workflow. Undeclared capabilities are risky because they expand the attack surface, may encourage secret handling or account probing, and defeat policy decisions based on the declared narrow purpose.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding

The skill advertises runnable scripts that can read environment variables, access files, invoke shell tools, and make network requests, but it does not declare any explicit tool scope or permission boundaries. In an agent environment, this creates an over-privileged integration surface where the skill may be invoked without clear least-privilege constraints or user visibility into what resources it can access.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The activation list includes generic phrases such as “服装视频” and broad conditions like requests in advertising, marketing, social media, or wanting to turn a reference case into original content. These descriptions are expansive enough to overlap with many ordinary content-creation requests, making it unclear when this skill should activate instead of a more general video or marketing skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The default prompt requires a "production-ready Chinese workflow," which imposes a specific language choice regardless of user preference. This is a natural-language policy concern unless the skill explicitly offers language selection or clearly justifies a Chinese-only 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 with no trigger scope, guardrails, or user-confirmation constraints, which can cause the agent to auto-select this skill in loosely related contexts. Because this skill is designed to generate runnable AI-HIVE commands and production workflows for media generation, unintended invocation could trigger sensitive content handling, external API preparation, or policy-bypassing assistance without sufficiently explicit user intent.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/edit_video.py (reported line 20)May include surrounding context.

python
def run(args: list[str]) -> None:
    print(" ".join(args))
    subprocess.run(args, check=True)


def main() -> None:

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module is presented and implemented as a broad, general-purpose AI Hive client with chat, image, video, upload, model enumeration, and account utilities, which exceeds the declared scope of an avatar try-on video skill. Scope expansion increases attack surface and enables unintended uses through this skill wrapper, weakening policy enforcement tied to the advertised purpose.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The top-level documentation advertises a generic AI capability tool, contradicting the specialized try-on video intent in the skill metadata. This mismatch is dangerous because reviewers, orchestrators, or downstream tooling may trust the manifest while the implementation exposes a much wider set of functions.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Exposing text chat in a skill intended for avatar try-on video creates an unnecessary capability channel that can be repurposed for unrelated or policy-bypassing tasks. Even if not directly invoked by the skill entrypoint, shipping this functionality in the same tool broadens reachable behavior and makes misuse easier if the script is reused or called differently.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Generic image generation is outside the declared try-on video purpose and permits the skill bundle to be used for unrelated content creation. This undermines least-privilege design and can bypass higher-level restrictions that assume the skill only produces controlled try-on video outputs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The description specifies invocation using Chinese phrases only and does not indicate that other languages are accepted or that the language choice depends on user preference. This creates a locale-specific constraint in the natural-language guidance without opt-in or justification as a region-limited skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The natural-language strings in the file, including the module docstring, CLI description, skill name, and search aliases, are entirely Chinese and present the skill identity in that locale without any opt-in or language-selection mechanism. The policy specifically calls for flagging language or locale constraints when the skill forces a specific language without user choice.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The module docstring and all user-facing help text are written in Chinese, presenting the skill as a general-purpose tool rather than a region- or locale-specific one. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is clearly documented and justified.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

Commands for querying user wallet/account info and listing all available models are convenience capabilities for a generic platform client, not an obvious requirement of a focused try-on-video skill. While useful operationally, they are broader than the manifest’s end-user purpose and reveal platform capabilities outside the skill’s declared function.

Content

No source excerpt is available for this finding.