Back to skill

Security audit

用户晒单视频自动混剪|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed AI-HIVE video workflow, but it handles API credentials and local video outputs in ways that need review before installation.

Install only if you trust the AI-HIVE endpoint configuration and will keep API keys out of logs and shared configs. Do not set --base-url or AI_HIVE_BASE_URL unless you control the server and understand that the API key will be sent there. Use fresh output filenames or backups when running the ffmpeg helper because it overwrites existing outputs by default.

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 (3)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/videogen.py:91
Finding

Bearer API Key Can Be Sent to an Arbitrary Server

Content
View full analysis
` header. Allowing plaintext HTTP also exposes the credential to network interception. This is not merely server-side request forgery: the principal risk is credential disclosure because the same sensitive authorization header is automatically sent to any configured destination. ### Attack Path 1. An attacker modifies the victim's environment, configuration file, wrapper script, or copied command to set: ```bash export AI_HIVE_BASE_URL=http://attacker.e ...[truncated 977 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/edit_video.py:62
Finding

Video Editing Commands Silently Overwrite Existing Output Files

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
scripts/videogen.py:7
Finding

Third-Party Dependency Is Installed Without Version or Integrity Pinning

Content
View full analysis
Remediation
View remediation
``` 2. Generate and verify cryptographic hashes for all transitive dependencies, and install with: ```bash python3 -m pip install --require-hashes -r requirements.txt ``` 3. Use a lockfile-producing dependency management tool to make transitive resolution reproducible. 4. Recommend installation in a dedicated virtual environment rather than the system Python environment. 5. Document the trusted package index and avoid silently inheriting unknown index settings in controlled deployment workflows. 6. Periodically review pinned versions for security advisories and update them through a tested dependency-review process. ]]>
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
Findings (15)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

This variant of the mismatch is more security-relevant because it suggests undeclared generic AI API, model enumeration, wallet/user-info, and image generation capabilities beyond the stated narrow montage purpose. Undeclared capabilities can bypass user expectations and review controls, especially when combined with network access and potential handling of billing-related or account-related data.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

This variant of the mismatch is more security-relevant because it suggests undeclared generic AI API, model enumeration, wallet/user-info, and image generation capabilities beyond the stated narrow montage purpose. Undeclared capabilities can bypass user expectations and review controls, especially when combined with network access and potential handling of billing-related or account-related data.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

This variant of the mismatch is more security-relevant because it suggests undeclared generic AI API, model enumeration, wallet/user-info, and image generation capabilities beyond the stated narrow montage purpose. Undeclared capabilities can bypass user expectations and review controls, especially when combined with network access and potential handling of billing-related or account-related data.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding

The skill advertises executable workflows involving environment variables, local file handling, network access, and shell-driven tooling, but it does not declare any explicit tool scope or permissions boundaries. In a skill that references API keys, ffmpeg, and outbound API calls, missing scope declarations increases the chance of overbroad execution, unintended data exposure, or unsafe tool invocation by an agent runtime.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The listed activation phrases include very broad terms such as “用户内容” and “品牌广告,” and the usage conditions also include generic business contexts like marketing, social media, and delivery scenarios. These phrases are common in ordinary discussion and may cause the skill to activate in situations far outside the intended montage workflow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The default prompt hard-codes output in Chinese regardless of the user's language preference, which can override user intent and reduce transparency or usability for non-Chinese-speaking users. In a security-sensitive workflow, forced language can also make review and operator validation harder, increasing the chance that harmful or noncompliant instructions pass unnoticed.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The user-facing description, argument parser description, skill name, and search aliases are written entirely in Chinese, which imposes a specific language/locale in the skill's natural-language interface. There is no indication that users can opt into another language or that this locale restriction is documented as a justified region-specific requirement.

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 file is presented as a specialized customer-review video montage skill, but the implementation is a generic AI client exposing text chat, image generation, model enumeration, upload, and task operations. This scope mismatch weakens least-privilege expectations for downstream agents and users, making it easier to invoke capabilities unrelated to the declared business purpose and increasing the chance of policy bypass or unintended data handling.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The module docstring explicitly describes a generic AI capability tool, contradicting the specialized montage branding in the skill metadata. This inconsistency is dangerous because agents and reviewers may apply the wrong trust assumptions, underestimating the real breadth of actions available through the script.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This code file contains user-facing natural-language descriptions and prompts entirely in Chinese, including the module docstring and later CLI guidance, but does not offer an opt-in language choice. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is explicitly justified.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Exposing user-info and wallet-balance retrieval is unnecessary for a montage-generation skill and reveals account metadata beyond the minimum needed to complete the task. In an agent setting, this can become an information disclosure path, especially if another component can trigger the command or surface the response to users/logs.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The generic text chat endpoint allows arbitrary prompts and optional system instructions, which is outside the declared montage-only scope. In practice, this broadens the skill into a general-purpose model proxy, enabling unintended content generation, policy circumvention, or processing of unrelated sensitive data through a skill that users may trust as narrowly scoped.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

Standalone image generation is not required for customer-review video montage assembly and materially expands the capability surface of the skill. Because the skill is marketed as a specialized montage tool, this hidden extra function can be used to generate unrelated synthetic assets or bypass stricter controls expected on a purpose-limited workflow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The skill name and description are presented entirely in Chinese and the invocation examples are Chinese-specific, with no indication that users may choose another language or locale. Under the policy, a fixed language expectation without opt-in can be a natural-language locale constraint.

Content

No source excerpt is available for this finding.