skills/freestylefly/wesight/seedance/Gen Agent Trust Hub

seedance

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-provided prompts that are sent to an external AI model without sanitization, creating a surface for indirect prompt injection.\n
  • Ingestion points: Prompt text received via the --prompt argument in generate_video.js.\n
  • Boundary markers: None; user input is passed directly to the model API.\n
  • Capability inventory: The skill performs file reads (fs.readFileSync, cat, Get-Content) and network requests (https.request) to Volcengine servers.\n
  • Sanitization: No validation or filtering is applied to the prompt content before transmission.\n- [COMMAND_EXECUTION]: The skill uses platform-specific entry scripts (.sh and .ps1) to resolve the Node.js environment and launch the core generation logic. These scripts also include a feature to read local file contents if an argument is prefixed with @.\n- [EXTERNAL_DOWNLOADS]: The script makes network connections to ark.cn-beijing.volces.com (Volcengine) to submit generation tasks and download the resulting MP4 video files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 05:16 AM
Security Audit — agent-trust-hub — seedance