Back to skill

Security audit

爆款视频镜头节奏拆解|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is a mostly disclosed video workflow helper, but it needs review because it handles API keys and can send them to a user-configurable API server.

Review before installing. Use only the official AI-HIVE endpoint unless you fully trust a custom server, avoid putting API keys in shared logs or commands, confirm any billable generation before running it, and use an isolated Python environment with reviewed dependencies.

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

API Key Can Be Forwarded to an Arbitrary Server Through a Configurable Base URL

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
scripts/videogen.py:7
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis
Remediation
View remediation
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 (20)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

There is a material description-behavior mismatch. The declared purpose presents a substantive video-analysis/editing workflow tool, potentially with external AI generation support. The supplied code only builds a static JSON brief from command-line arguments. Its primary purpose is metadata templating, not media analysis or editing assistance. While some field names and aliases align with the theme, the implemented behavior is far narrower and different from the advertised capability.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The declared description presents an analysis-and-planning skill for viral-style video restructuring, focused on editorial reasoning about timeline rhythm and possibly AI-assisted asset generation. The supplied code instead implements a small ffmpeg wrapper for straightforward media editing utilities. While both relate broadly to video, the primary purpose is materially different: the code edits files directly but does not analyze shots, infer pacing, explain transitions, create B-roll recommendations, or orchestrate any AI-HIVE generation workflow. There are no suspicious extra capabilities beyond the basic ffmpeg operations, but the implemented behavior substantially underdelivers and differs from the declared functionality, so this is a clear mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The description promises a domain-specific analysis tool for breaking down editing rhythm and reconstructing video pacing from supplied authorized footage. However, the code does not analyze video structure, detect shots, compute rhythm peaks/valleys, infer transition logic, produce B-roll recommendations, or generate a recut plan. Instead, it is an API wrapper around AI Hive with generic capabilities: chat, image generation, video generation, uploads, polling, downloading, model lookup, and user-info retrieval. The skill-configured layer maps prompts and optional reference media into text-to-video / image-to-video / reference-to-video generation calls, which is materially different from analytical video-editing assistance. There are also undeclared capabilities such as text chat, image generation, and account info access present in the codebase. Therefore the declared purpose does not accurately represent the actual behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The file implements a broad AI-Hive client for chat, image generation, video generation, model enumeration, uploads, and account queries, while the declared skill is narrowly about shot-rhythm analysis and re-edit planning. This scope mismatch materially increases the accessible attack surface and enables unrelated capabilities to be invoked under the guise of a specialized skill, which violates least privilege and can be abused for unauthorized content generation or data access.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The module explicitly presents itself as a 'general AI capability invocation tool,' directly contradicting the manifest's specialized shot-rhythm-analysis purpose. This kind of misleading packaging is dangerous because it obscures actual capabilities from reviewers and users, making overbroad permissions and hidden functionality easier to slip into production.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The primary skill path routes directly into video generation rather than analysis of an authorized source video's pacing, cuts, transitions, and timeline structure. In context, this is more dangerous because the manifest positions the skill as an analysis tool, but the code enables media synthesis, creating a misleading permission boundary that could be used for unapproved content creation or repurposing user-supplied assets beyond expected analysis.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill advertises executable workflows involving shell commands, environment variables, file access, and network/API usage, but it declares no explicit tool scope or permission boundaries. In an agent setting, this can lead to over-broad capability exposure, making it easier for downstream prompts or users to trigger filesystem, network, or shell actions that were never formally constrained or reviewed.

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 description and title are entirely in Chinese and present the skill as operating in that language, with no indication that users may choose another language. Under the policy rule, a skill should not implicitly force a specific language or locale unless it offers opt-in or clearly documents a justified regional constraint.

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 hard-codes Chinese output regardless of the user's language preference, which can override user intent and reduce transparency about what the agent is doing. While this is not a direct code-execution issue, forced language can impair review, consent, and safe use, especially when the skill may emit operational commands or workflow instructions the user cannot easily validate.

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 contexts the user did not clearly intend. Because this skill is designed to produce production-ready workflows and runnable AI-HIVE commands, unintended activation could cause the model to steer users into external API usage, file handling, or cost-incurring generation flows without clear consent or relevance checks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The natural-language description and skill name are hardcoded in Chinese, with no option for the user to choose another language or locale. This can violate language/locale policy when a skill imposes a specific language without opt-in or justification.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The blueprint declares artifacts such as '差异化复刻方案', '逐镜脚本', and '生成提示词', which go beyond a passive analysis brief and enable downstream replication or remake workflows. In this skill context, that broadens the tool from analysis into content reproduction planning, increasing misuse risk for unauthorized imitation or policy-evasive ad/video generation.

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
94% confidence
Finding

The manifest describes a skill for analyzing authorized videos and producing outputs like shot-duration tables, rhythm peaks/valleys, transition logic, B-roll lists, and recut plans. This helper instead implements concrete ffmpeg edit actions such as trimming, aspect-ratio transforms, muting, loudness normalization, and concatenation, which are execution-oriented editing behaviors rather than the described analytical deliverables.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The code exposes a user-info endpoint that returns account and wallet-related data unrelated to shot-rhythm analysis. Even if authenticated access is required, surfacing billing/account information in a content-analysis skill broadens data exposure and may leak sensitive operational details to users or downstream agents that should never need them.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The generic text chat function allows arbitrary model prompting that is not necessary for a dedicated shot-rhythm analyzer. This creates an unsanctioned general-purpose LLM interface inside a narrowly scoped skill, allowing policy bypass, prompt abuse, or unintended processing of user/media-derived data outside the promised workflow.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Standalone image generation is outside the declared analysis workflow and adds content-creation capability unrelated to timeline decomposition. In this skill context, that mismatch makes the issue more suspicious because users may authorize media analysis while the implementation can instead synthesize unrelated assets through the same trusted integration.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

The docstring says the script will create a deterministic production brief for shot-rhythm analysis, suggesting a bounded planning document. However, the actual payload includes '差异化复刻方案', '逐镜脚本', and '生成提示词和验收清单', which goes beyond a narrow analysis brief into prescriptive generation and replication-oriented work.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This code creates parent directories and writes a JSON file to a user-specified path, which is a filesystem-modifying operation. There is no confirmation prompt, warning message, or explanatory comment/docstring near the write path that clearly discloses this side effect to the user.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

This Python file contains user-facing natural-language text entirely in Chinese, including the module description and many CLI messages, while presenting itself as a general-purpose AI skill. Because the skill does not offer locale selection or document that it is intentionally Chinese-only, it may violate language/locale policy requiring user choice or explicit justification.

Content

No source excerpt is available for this finding.