Back to skill

Security audit

口播视频去废话精剪|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is not clearly malicious, but it needs review because it uses API credentials and remote media uploads while overstating its automatic video-editing capability and allowing weakly scoped API endpoints.

Install only if you are comfortable giving this skill an AI-HIVE API key and uploading selected media to AI-HIVE. Use the official HTTPS endpoint, avoid custom base URLs unless you fully control them, use a scoped/revocable API key, and do not expect the scripts alone to automatically remove pauses, generate subtitles, or produce editorial timecodes without additional manual or agent work.

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

Unrestricted API Base URL Can Redirect Credentials and Media to an Attacker-Controlled Server

Content
View full analysis
Remediation
View remediation
str: parsed = urlparse(value) if parsed.scheme != "https": raise SystemExit("The API endpoint must use HTTPS.") if parsed.hostname not in TRUSTED_API_HOSTS: raise SystemExit("The API endpoint is not trusted.") if parsed.username or parsed.password: raise SystemExit("Credentials in the API URL are not permitted.") return value.rstrip("/") ``` 4. Validate upload responses: - Require HTTPS upload URLs. - Allowlist documented object-storage domains. - Reject unexpected HTTP methods. - Reject redirects or revalidate every redirect destination. - Do not forward headers returned by an untrusted API without filtering. 5. Protect the configuration file: - Continue enforcing mode `0600`. - Reject symbolic links when reading or writing sensitive configuration where practical. - Warn if the stored endpoint differs from the official production endpoint. 6. Use scoped, revocable API keys and provide clear key-rotation instructions in case endpoint configuration is compromised. ]]>

T08 · Insecure Dependencies

Warning
Location
scripts/videogen.py:7
Finding

Unpinned Runtime Dependency Installation Creates a Supply-Chain Risk

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (11)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

There is a material description-behavior mismatch. The declared purpose describes a full spoken-video refinement/editing skill, including analysis of source media/transcripts and possibly AI-HIVE-backed generation workflows. The supplied code does none of that; it simply parses CLI arguments such as project, audience, goal, platform, and format, then writes a static JSON draft blueprint with preset fields like aliases, required facts, and artifacts. This is not a supporting implementation detail of the described functionality—it is a substantially different and much narrower capability centered on planning metadata creation. Therefore the description does not accurately represent the actual code behavior.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The declared description promises a higher-level automated口播精剪 system aimed at speech-video refinement, including transcript-aware editing decisions and optional external AI-HIVE API orchestration. The supplied code instead exposes only basic command-line ffmpeg wrappers for manual media operations: probe, trim, aspect conversion, mute, loudness normalization, and concat. These are related to video processing in a broad sense, but they do not implement the core declared behavior of automatic speech cleanup, semantic restructuring, subtitle output, or remote generation workflow. Because the primary purpose and key capabilities are materially different and much narrower than described, this is a clear description-to-behavior mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The description claims a specialized spoken-video editing workflow focused on removing filler/silence/repetition while preserving speaking logic and tone, and delivering editorial artifacts such as retained segments, deletion suggestions, timestamps, pacing rearrangement, subtitles, and vertical edited outputs. The supplied code does not implement any of that editing logic. There is no video parsing, silence detection, transcript ingestion, segmentation, timestamp generation for cuts, subtitle generation, or assembly of an edited final cut. Instead, the code primarily exposes generic AI Hive API operations: text chat, image generation, video generation, model lookup, account info, media upload, and asynchronous task polling/download. The skill-specific layer merely constrains calls to certain AI video generation models and forwards prompt/reference inputs to a video-generation endpoint. While the declared description mentions some AI-HIVE plumbing that the code indeed uses (config lookup, upload, pricing snapshot, routing modes, polling, download), those are only infrastructure details and do not make the primary purpose match. The actual primary behavior is a general/generative AI client with a video-generation wrapper, materially different from a talking-video silence-cut/fine-editing skill.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The implementation materially differs from the advertised purpose: instead of performing silence-cutting or speech-editing, it provides a generic AI-Hive client for chat, image, and video generation plus media upload/download. In a skill ecosystem, this kind of capability mismatch is dangerous because users or orchestrators may grant the tool sensitive media and broad trust under a narrow editing use case, while the code can route that data to unrelated remote generation APIs and perform unexpected operations.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill documents executable workflows that use environment variables, local files, network access, and shell-driven tooling, but it does not declare any explicit tool scope or permission boundaries. In an agent environment, that mismatch weakens least-privilege controls and can lead to unintended file access, network calls, or command execution if the skill is invoked broadly or interpreted too permissively.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The skill name, description, and heading are entirely framed in Chinese and present the skill as activated by Chinese phrases such as '口播精剪' and '去废话'. There is no explicit statement that users may interact in other languages or choose their preferred locale, which can constitute a language/locale policy violation under the stated rule.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The natural-language instruction says to produce a "production-ready Chinese workflow," which imposes a specific language choice by default. This is a locale/language policy concern because the file does not offer the user a choice or indicate that Chinese is required for a justified region-specific purpose.

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:

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The top-level documentation and CLI describe a generic AI generation tool and video generation skill, which conflicts with the skill's declared purpose of talking-video silence trimming. This inconsistency can mislead reviewers, users, and higher-level agents about what the tool actually does, increasing the chance of unauthorized data handling or misuse of a broadly capable backend under a narrow, trusted label.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The module docstring and all user-facing CLI guidance are written as Chinese-only instructions, presenting the skill as a general-purpose tool rather than a region-specific or explicitly Chinese-only skill. Under the policy, forcing a specific language without user opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

This code hard-codes the skill name and argument parser description in Chinese ("口播视频去废话精剪|AI-HIVE") with no indication that users may choose another language or locale. Under the stated policy, forcing a specific language without opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.