Back to skill

Security audit

抖音内容提取总结

Security checks for vulnerabilities and agentic risk

Overview

This video transcription skill mostly does what it claims, but it under-discloses risky network behavior and makes privacy claims that are not verifiable from the artifact.

Install only if you are comfortable sending selected videos, video URLs, and prompts to the Guaikei service. Avoid using it on confidential meetings, private network URLs, cloud metadata URLs, internal services, or very large/untrusted links until the publisher adds explicit URL allowlisting, private-address blocking, size/time limits, and a verifiable privacy/retention policy.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/utils/validator.js:3
Finding

Server-Side Request Forgery Through Unrestricted Video URLs and Redirects

Content
View full analysis
{ if (this.__isRequireRedirect(response)) { redirectCount++; if (redirectCount > this.__opts.maxRedirects) { const err = new Error("Too many redirects"); this.__setState(this.__states.FAILED); this.emit("error", err); return reject(err); } const redirectedURL = /^https?:\/\//.test(response.headers.location) ? response.headers.location : new URL(response.headers.location, url).href; ``` `scripts/utils/download.js:479-491`: ```js if (this.__isRequireRedirect(response)) { this.__redirectCount++; if (this.__redirectCount > this.__opts.maxRedirects) { const err = new Error("Too many redirects"); this.__setState(this.__states.FAILED); this.emit("error", err); return reject(err); } const redirectedURL = /^https?:\/\//.test(response.headers.location) ? response.headers.location : new URL(response.headers.location, this.url).href ...[truncated 2781 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/utils/download.js:34
Finding

Unbounded Remote Downloads Allow Disk, Bandwidth, and Execution-Time Exhaustion

Content
View full analysis
= totalSize) ) { this.emit("skip", { totalSize: this.__total, fileName: this.__fileName, filePath: this.__filePath, downloadedSize: downloadedSize, }); this.__setState(this.__states.SKIPPED); return resolve(true); } } this.__fileStream = fs.createWriteStream(this.__filePath, {}); } else { this.__fileStream = fs.createWriteStream(this.__filePath, { flags: "a" }); } this.emit("download", { fileName: this.__fileName, filePath: this.__filePath, totalSize: this.__total, isResumed: this.__isResumed, downloadedSize: this.__downloaded, }); this.__retryCount = 0; this.__isResumed = false; this.__isRedirected = false; this.__setState(this.__states.DOWNLOADING); this.__statsEstimate.time = new Date ...[truncated 2908 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (23)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

声明描述的是一个视频转写与文案生成技能,核心能力应包括接收视频输入、调用转写/LLM 服务、清洗文本并输出结构化内容。而实际代码仅实现了通用命令行参数解析与帮助文本生成:识别 flag、处理默认值、校验必填项、支持 --help 和 --flag=value 等。该代码片段的主要目的与声明严重不符,且没有体现任何与视频、音频、字幕、文本生成相关的行为,因此属于明显的描述-行为不匹配。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

代码与“视频转文字/总结生成”没有直接关系。该代码块的核心功能是下载任意 URL 对应的远程资源到本地目录,并管理下载生命周期;它没有任何视频解析、音频提取、语音识别、字幕提取、大模型调用、文本清洗、总结或文案生成逻辑。虽然下载远程视频链接可能是整条技能链路中的一个辅助步骤,但当前代码块本身体现的是独立且未声明的网络下载与文件写入能力,因此与声明的主要用途存在明显不一致。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

从该代码片段本身看,它的核心职责是下载远程文件并显示下载进度,属于基础工具层实现。虽然声明中提到支持抖音、小红书等平台视频链接,这在整体技能中可能需要下载步骤作为前置支持,但当前片段没有任何视频解析、音视频处理、语音转写、文本清洗、结构化生成或模型调用逻辑。因此,就“该代码片段实际做什么”而言,与声明的主要用途不一致,应判定为描述与行为不匹配。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

声明描述的是一个完整的视频转文字与内容加工技能,但提供的代码片段只实现了一个辅助函数:通过 fs 读取 package.json 并返回包名,且使用了缓存优化。该行为与视频文件处理、链接解析、模型调用、文本清洗、总结改写等声明能力均无直接对应关系。虽然这可能是某个内部工具函数,但就当前提供的代码块而言,其实际行为与声明用途明显不匹配。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

声明描述的是一个视频转文字与文案生成技能,但提供的代码片段并未展示任何与视频输入、音视频解析、语音转写、LLM 总结改写或平台链接处理相关的行为。该代码的实际用途是验证 token 是否符合格式要求,并在无效时打印警告、推广网址和客服微信等信息。这属于与声明主功能明显不一致的实现片段,且其主要行为与技能描述无直接对应关系,因此应判定为不匹配。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The entire skill description, examples, contact flow, and onboarding instructions are presented only in Chinese, including WeChat-based token acquisition and prompt examples. This effectively imposes a specific language/locale on users without any opt-in or alternative language support being documented.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The README makes strong privacy and deletion assurances about encrypted temporary storage and automatic destruction of uploaded videos, but this file provides no technical evidence, contractual control, or verifiable mechanism to enforce those claims. For a skill that handles potentially sensitive video content, unverifiable data-handling promises can mislead users into uploading confidential material under false assumptions.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The claim that uploaded materials will not be used for model training or external distribution is a trust and data-governance assertion with no substantiation in the skill artifact. In this context, users may submit proprietary business recordings, interviews, or meetings, so unsupported promises can create privacy, compliance, and confidentiality risk if backend practices differ from the README.

Content

No source excerpt is available for this finding.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · README.md (reported line 99)May include surrounding context.

md
4. 同时传入文件路径与任务ID,优先执行 `--id`,忽略 `--file`
5. 无自定义 prompt 时,默认完整转录视频全部文字

| 用户自然语言指令                                         | 自动生成命令                                                                                                |
| -------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| 视频提取 https://example.com/video.mp4 中的文字          | `node scripts/video2text/index.js --file "https://example.com/video.mp4"`                                   |
| 把本地 /path/to/your/video.mp4 改成小红书风格的文案      | `node scripts/video2text/index.js --file "/path/to/your/video.mp4" --prompt "改写成小红书风格的文案"`       |

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The trigger conditions are very broad and cover many generic content-processing intents, increasing the chance the skill is invoked for unrelated user requests. In an agent environment, overbroad auto-invocation can cause unintended transmission of user-provided local files, remote URLs, prompts, or sensitive content to a third-party service.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The usage guidance prioritizes this skill for a wide range of loosely related tasks without strong boundaries or exclusions. That makes accidental invocation more likely, which is risky here because the skill sends user content to an external service and encourages automated natural-language-to-command mapping.

Content

No source excerpt is available for this finding.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 125)May include surrounding context.

md
当用户用自然语言下达指令时,按以下映射生成命令,保证识别与执行一致:

| 用户自然语言指令                                     | 生成的命令                                                                                                  |
| ---------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| 提取 https://example.com/video.mp4 里的文字          | `node scripts/video2text/index.js --file "https://example.com/video.mp4"`                                   |
| 总结这个视频的核心观点 https://example.com/video.mp4 | `node scripts/video2text/index.js --file "https://example.com/video.mp4" --prompt "总结这个视频的核心观点"` |

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This utility’s comments and user-facing error/help text are consistently hard-coded in Chinese, indicating a fixed language/locale choice. The file does not offer any mechanism for user language selection or document that the tool is intentionally limited to a Chinese-only context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Multiple thrown errors and generated help text use only Chinese phrases such as 参数缺少值, 未识别的选项, 用法, 选项, and 示例. Because the file exposes these strings to end users and provides no opt-in or alternative locale, it enforces a specific language choice.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The default options enable removeOnStop: true and removeOnFail: true, and later code unlinks the downloaded file during stop and error handling. Although the behavior is implemented correctly, this file provides no confirmation prompt, user-facing disclosure, or warning comment/docstring that stopping or failed downloads may delete files.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This JavaScript file contains natural-language strings and comments such as the doc comment at L24-L25 and multiple status messages below that are written only in Chinese. The file does not indicate that the skill is region-specific or provide any user opt-in for language selection, which can violate organizational language/locale policy.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This code emits all user-visible error strings in Chinese, including authentication and network failures, with no indication that the skill is intentionally limited to Chinese-speaking users. The stated policy flags language/locale constraints that are imposed without user opt-in or clear justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The string literals shown to users are entirely in Chinese and the file does not offer any language selection, fallback, or indication that the skill is intentionally limited to a Chinese-speaking audience. This creates a natural-language locale policy issue because the skill imposes a specific language on users without opt-in or documented justification.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The comment makes a privacy and data-handling promise that uploaded videos are only used for transcription and automatically deleted afterward, but this module contains only upload logic and no deletion or retention enforcement. In a skill that processes potentially sensitive user videos, this kind of mismatch can lead to unintended retention of personal or confidential data and creates a false assurance for users and integrators about lifecycle controls.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This code presents argument descriptions, examples, and error/help text only in Chinese, which effectively forces a specific language for users. The file does not offer a language selection mechanism or indicate that the skill is intentionally limited to a Chinese-speaking context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This code sends a user-provided video URL to the /api/video/id endpoint and later sends a prompt to /api/video/text, which may disclose user content to a remote service. Although the functions have technical docstrings, there is no visible confirmation prompt, warning, or user-facing disclosure in this file about transmitting potentially sensitive media metadata or instructions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The banner string is hard-coded in Chinese, which indicates the skill presents a fixed language experience. Under the policy criteria, forcing a specific language without user opt-in or a documented justification is a natural-language policy concern.

Content

No source excerpt is available for this finding.