sora
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads video, thumbnail, and spritesheet assets from OpenAI's official API endpoints. These operations are performed via the authorized OpenAI Python SDK.
- [COMMAND_EXECUTION]: The skill utilizes a local Python script (
scripts/sora.py) to execute API commands. Instructions guide the agent to use standard tools likeuvor the Python interpreter for execution. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided prompt files and JSONL batch files, which constitutes an ingestion surface for untrusted data.
- Ingestion points: Reads external prompt text from files via the
--prompt-fileargument and batch job definitions from JSONL files via the--inputargument inscripts/sora.py. - Boundary markers: Prompt content is passed as an argument to the CLI but is not encapsulated with specific safety delimiters within the script logic.
- Capability inventory: The skill can perform network operations to the OpenAI API and write generated files to the local file system.
- Sanitization: The skill does not perform content-based sanitization or filtering of the prompt text before it is sent to the generative API.
Audit Metadata