capcut-ai-video-editor

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill is configured to upload local video files and metadata to the external API at https://mega-api-prod.nemovideo.ai/api/upload-video/. Although essential for the skill's purpose, this behavior involves transmitting user data to a third-party service provider.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes Server-Sent Events (SSE) from the remote backend and maps specific text commands to agent actions (e.g., "Export button" triggers the export workflow). This creates an attack surface where a compromised backend could control the agent's behavior.
  • Ingestion points: Local file data for upload and the SSE response stream from the backend.
  • Boundary markers: The skill does not implement boundary markers or instructions to differentiate data from executable commands within the backend's response.
  • Capability inventory: Includes network operations (POST/GET), file read access for media uploads, and management of authentication tokens.
  • Sanitization: No validation or sanitization is performed on instructions received from the remote API.
  • [EXTERNAL_DOWNLOADS]: The skill retrieves authentication tokens and session states from the non-whitelisted domain nemovideo.ai.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 11:32 PM
Security Audit — agent-trust-hub — capcut-ai-video-editor