Back to skill

Security audit

爆款视频字幕与文案拆解|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is not clearly malicious, but it can use an AI-HIVE API key to upload, generate, poll, and download media, which is broader and more sensitive than the analyzer branding suggests.

Review this before installing. Use a limited AI-HIVE API key, avoid custom base URLs unless you fully control the endpoint, and require explicit confirmation before any upload or generation because those actions may expose media and consume paid credits. Do not paste real API keys into chat logs or project files.

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

Warning
Location
scripts/videogen.py:101
Finding

Bearer API Key Can Be Forwarded to an Untrusted Custom Base URL

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
scripts/videogen.py:6
Finding

Third-Party Dependency Is Installed Without Version or 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
96% confidence
Finding

This mismatch is more security-relevant because the skill is presented as a subtitle/copy analysis assistant while apparently enabling broader AI-HIVE account and media operations such as account info retrieval, model/pricing queries, uploads, and downloads. If users invoke it expecting low-risk text analysis, the hidden expansion into networked account/resource actions can lead to unintended data disclosure, billing exposure, or external content transfer.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

This mismatch is more security-relevant because the skill is presented as a subtitle/copy analysis assistant while apparently enabling broader AI-HIVE account and media operations such as account info retrieval, model/pricing queries, uploads, and downloads. If users invoke it expecting low-risk text analysis, the hidden expansion into networked account/resource actions can lead to unintended data disclosure, billing exposure, or external content transfer.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

This mismatch is more security-relevant because the skill is presented as a subtitle/copy analysis assistant while apparently enabling broader AI-HIVE account and media operations such as account info retrieval, model/pricing queries, uploads, and downloads. If users invoke it expecting low-risk text analysis, the hidden expansion into networked account/resource actions can lead to unintended data disclosure, billing exposure, or external content transfer.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file is presented as a caption/copy analysis skill, but the implementation is a general AI Hive client for chat, image, and video generation. This mismatch can mislead users and higher-level agents into invoking media generation, uploads, browser opening, polling, and file download behaviors they did not intend to authorize, expanding the skill's effective capability surface.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The embedded configuration brands the tool as a viral-caption/copy analyzer but binds only to text-to-video, image-to-video, and reference-to-video models. This is a capability deception issue: consumers may trust the branding and hand over prompts or assets expecting analysis, while the code actually submits them to media-generation services.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill-specific entrypoint routes the primary generate action into video generation rather than subtitle/copy analysis. In an agent ecosystem, this is dangerous because orchestration may select the skill based on its advertised analysis purpose, yet execution triggers remote content generation and associated side effects.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill advertises and demonstrates capabilities that imply network access, shell execution, file I/O, and environment-variable use, but it declares no explicit tool scope or permission boundary. That creates an authorization ambiguity where a caller or host may grant broader capabilities than users expect, increasing the chance of unintended external calls, local file access, or secret handling during skill execution.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The default prompt explicitly instructs the skill to produce a "production-ready Chinese workflow," which imposes a language choice regardless of user preference. This is a natural-language policy concern because the file does not offer an alternative language option or document that the locale restriction is intentional and user-approved.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

allow_implicit_invocation: true lets the platform auto-activate this skill without a tightly scoped trigger boundary. Because the skill is capable of generating production-ready workflows and runnable AI-HIVE commands, broad implicit activation increases the chance of unintended execution paths, overbroad prompt capture, or misuse in contexts the user did not clearly authorize.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill’s natural-language identifiers and defaults are entirely in Chinese, including the description, argument description, skill name, and search aliases, with no indication that users can choose another language. This can violate language/locale policy when a skill implicitly constrains interaction to one language without explicit opt-in or documented justification.

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:

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest describes a skill focused on analyzing and rewriting subtitles, spoken copy, CTA text, and optionally using AI-HIVE APIs for asset-generation workflows. This script instead performs direct local video-editing operations such as trimming, aspect-ratio conversion, muting audio, loudness normalization, and concatenation via ffmpeg, which are production media-processing capabilities not justified by the stated text/caption-analysis role.

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 is written entirely in Chinese and presents the tool as a general-purpose skill, while user-facing prompts and guidance throughout the file are also fixed in Chinese. For a broadly scoped skill, this creates a language/locale constraint without offering the user any language choice or documenting that the skill is region-specific.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The parser description claims subtitle/copy decomposition while the code labels itself as a raw video skill and exposes video-generation commands. Although primarily a trust and integrity issue rather than code execution, this inconsistency increases the chance of unsafe invocation and unintended data transfer in automated tool selection contexts.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

This code writes a JSON file to a user-supplied path and will overwrite existing content at that location, but there is no confirmation prompt, warning message, or explanatory comment/docstring disclosing that behavior. Because file modification is a safety-relevant operation in code files, users should be clearly informed before or during execution.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The skill creates an output directory and downloads generated media files into it automatically. Although the code shows download progress, it does not clearly warn the user beforehand in a docstring/comment/help text that running generation will write files under the default Downloads path.

Content

No source excerpt is available for this finding.