sora
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is configured to run a bundled Python script (
scripts/sora.py) to handle video generation, polling, and management tasks. - [EXTERNAL_DOWNLOADS]: The skill facilitates the download of video assets, thumbnails, and spritesheets from OpenAI's API endpoints as part of its primary video generation workflow.
- [PROMPT_INJECTION]: The skill ingests untrusted user prompts for video generation, which represents an indirect prompt injection surface. Mitigation is provided through structured prompting:
- Ingestion points: User-provided text prompts for video creation or remixing.
- Boundary markers: Detailed templates with field headers (e.g., 'Primary request:', 'Action:', 'Constraints:') are used to wrap and isolate user input.
- Capability inventory: Script execution, file management in temp directories, and network requests to OpenAI.
- Sanitization: Instructions require reformatting input into a fixed specification and advise against inventing details not provided by the user.
Audit Metadata