Back to skill

Security audit

Seedance 2.0 Video Gen

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent Markdown-only helper for preparing Seedance/EvoLink video-generation API requests, with no hidden execution, persistence, or local data access.

Before using generated API calls, confirm you are comfortable sending the prompt and any referenced media URLs to EvoLink/Seedance and use your own SEEDANCE_API_KEY. The skill would benefit from clearer privacy and confirmation language, but the inspected artifacts are proportionate to a video-generation API helper.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (10)

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger phrases are very broad and map to common user intents like '生成视频' and '制作视频', which can cause the skill to activate unexpectedly for ordinary conversations about video creation. Over-broad invocation increases the chance that user prompts, media references, or URLs are routed to this external-integration skill without clear user awareness.

Natural-Language Policy Violations

Medium
Confidence
88% confidence
Finding
The file sets `language: zh-CN`, and the surrounding content is entirely in Chinese, but it does not indicate that this is an optional default or that the skill is region-specific. Under the policy, forcing a specific language or locale without user opt-in is a natural-language policy concern.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill instructs collection and packaging of prompts plus image/video/audio URLs for submission to a third-party API, but it does not warn users that their content will be transmitted externally. This creates a meaningful privacy and consent risk, especially because uploaded media may contain sensitive, copyrighted, or identifying information.

External Transmission

Medium
Category
Data Exfiltration
Content
**接口地址:**
```
POST https://api.evolink.ai/v1/videos/generations
GET  https://api.evolink.ai/v1/tasks/{task_id}
```
Confidence
95% confidence
Finding
The skill explicitly targets an external API endpoint, meaning user prompts and potentially user-supplied media references are transmitted outside the local agent environment. In this context, external transmission is expected functionality, but it is still security-relevant because the skill does not pair it with consent, minimization, or trust-boundary warnings.

External Transmission

Medium
Category
Data Exfiltration
Content
**接口地址:**
```
POST https://api.evolink.ai/v1/videos/generations
GET  https://api.evolink.ai/v1/tasks/{task_id}
```

**认证:** `Authorization: Bearer $SEEDANCE_API_KEY`
Confidence
95% confidence
Finding
The task polling endpoint is part of the same third-party service boundary, so task identifiers and related request metadata are also disclosed externally. Although normal for this integration, the skill provides no warning about third-party processing or the implications of using a bearer-token-authenticated external service.

Natural-Language Policy Violations

Medium
Confidence
84% confidence
Finding
The file presents writing guidance under Chinese sections but repeatedly prescribes English-only prompt formulations such as camera, action, and audio descriptions. Because the template does not state that English is optional or required for a justified tool limitation, it effectively imposes a language preference without user opt-in.

Natural-Language Policy Violations

Low
Confidence
82% confidence
Finding
This markdown file is primarily written in Chinese, but several recommended prompt templates are presented only in English as normative examples. Because the skill content does not state that English is optional or required for a justified reason, it can be read as forcing a language choice without user opt-in.

Natural-Language Policy Violations

Low
Confidence
79% confidence
Finding
The case guidance continues to prescribe English prompt wording inside a Chinese document, which may imply a fixed output language. Under the language/locale policy check, this should either be optionalized or justified as model-specific guidance.

Natural-Language Policy Violations

Low
Confidence
77% confidence
Finding
The product template is written entirely in English while the surrounding instructions are Chinese, and there is no language-choice note. This creates a mild policy issue because the document appears to standardize one language without explicit opt-in or locale justification.

Natural-Language Policy Violations

Low
Confidence
77% confidence
Finding
This section provides another default prompt template in English only, despite the rest of the file being in Chinese. Without an opt-in or a documented reason for the language constraint, it may violate the language/locale policy criterion.

Static analysis

No suspicious patterns detected.