assets
Warn
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions recommend installing and running the
@agent-native/coreandskillspackages from the npm registry usingnpxto set up the environment. - [COMMAND_EXECUTION]: The installation and connection steps involve executing shell commands (
npx @agent-native/core@latest skills add assets) which download and run code from the npm registry. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts to generate images and videos, representing a vulnerability surface.
- Ingestion points: User-provided strings passed as the
promptargument to tools such asopen-asset-picker,generate-image, andgenerate-video. - Boundary markers: None identified; user input is directly interpolated into generation requests.
- Capability inventory: The skill performs network requests to external generation services (
assets.agent-native.com) and writes the resulting media files (previewUrl,downloadUrl) to the local filesystem to provide inline previews. - Sanitization: No evidence of input validation or sanitization is present in the instructions for handling the user-supplied prompts.
Audit Metadata