other
Warning
- Location
SKILL.md:132- Finding
Guidance Facilitates Content-Moderation Evasion
- Content
View full analysis
- Remediation
View remediation
Security audit
seedance 2.0 guide
Security checks for vulnerabilities and agentic risk
Overview
This is mostly a video prompt guide, but it includes guidance that can help users evade face or content checks for identity-linked media generation.
Review this skill before installing if you use real people's images or likenesses. The Markdown-only package does not appear to compromise the local environment, but its prompt guidance should be revised to remove detector-aware face-obscuring and term-substitution advice and to require consent and compliance with the video platform's content policies.
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
Findings (1)
Vulnerability Patterns
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
- Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (4)
Ssd 4
Medium
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
Taken together, the stylization guidance, facial obscuring techniques, and descriptor substitution create a multi-step workflow for gradually transforming a likely sensitive face-generation request into one less likely to be blocked. This normalization of layered evasion materially increases the chance that users can circumvent safeguards for realistic or identity-linked visual generation.
- Content
Ssd 2
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The skill recommends composition choices like side profile, long shot, silhouette, backlighting, and obscured facial details specifically because they reduce facial visibility and notes that centered high-realism faces are 'the first place the system checks.' This moves beyond neutral cinematography advice into detection-avoidance tactics for generating otherwise restricted realistic face content.
- Content
Ssd 2
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
This section explicitly advises replacing realistic-face and skin-related terms with alternative wording such as 'CGI Digital Human' or 'Android shell' to avoid triggering safety checks while preserving the same intended visual outcome. That is semantic evasion guidance: it teaches users how to rephrase disallowed or sensitive requests so the underlying system is less likely to detect them.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
This file presents the skill documentation entirely in Chinese, which can constitute a language/locale policy violation when the skill does not explicitly offer language choice or explain that the locale restriction is required. Although there is a link to an English README, this file itself does not state that language selection is optional or user-driven.
- Content
Static analysis
No suspicious patterns detected.