HotBee Douyin Video Report

• Outcome: Pass
• Latest audit: Aug 9, 2026 · 2:48 AM UTC
• Version: 1.0.1

Overview

This skill is a disclosed HotBee/Douyin report generator that saves local analysis files from a user-provided public video link.

Install only if you are comfortable using your HotBee key and quota to process public Douyin links. Review generated reports, raw files, comments CSV/JSON, and transcripts before sharing them or committing them to a repository because they may contain public usernames, comments, timestamps, and location labels.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 95% confidence
• Finding: The skill instructs the agent to read an environment variable, access local files, write multiple output artifacts, and make outbound network requests, yet it does not declare permissions. This creates a trust and consent gap: a user or platform may not realize the skill can access sensitive local resources and external services, increasing the risk of unintended data exposure or over-broad execution.

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The default prompt tells the agent to use the skill for any provided Douyin link and to generate multiple output artifacts, but it does not constrain source validation, user authorization, data minimization, or safe handling of third-party content. In a skill that fetches external video metadata, captions, comments, and media-derived artifacts, broad invocation language increases the chance of over-collection, misuse on unauthorized targets, or unsafe downstream processing triggered from an unscoped request.

Natural-Language Policy Violations

Medium

• Confidence: 80% confidence
• Finding: The metadata is written entirely in Chinese and the default prompt prescribes Chinese-language behavior without indicating that language should follow user preference. This can cause the agent to ignore the user's requested language or hide important consent, privacy, or operational details from users who cannot read Chinese, which is a safety and transparency issue for a data-collection skill.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The script stores comment text, author nickname, inferred location/IP label, timestamps, and engagement metadata into local CSV/JSON/HTML artifacts without any consent prompt, minimization, or retention controls. Even though this is framed as analytics/report generation, it creates a privacy and compliance risk because third-party user data is being collected and persisted beyond transient processing.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The transcript feature sends a video URL to an external speech-to-text API endpoint, which shares user-requested content with a third party without explicit notice or confirmation. In this skill's context, users may expect local report generation, but the implementation discloses content to a remote service and may expose private, deleted-later, or access-controlled media references.

VirusTotal

1/65 vendors flagged this skill as malicious, and 64/65 flagged it as clean.

Static analysis

No suspicious patterns detected.