klingai-camera-control
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and scene descriptions for video generation, creating a surface for potential indirect prompt injection attacks.
- Ingestion points: User input is accepted through prompt fields in
SKILL.mdand processed via Python methods inreferences/camera-control-implementation.md. - Boundary markers: The instructions provide natural language guidelines for manual review and sandbox testing but do not implement technical delimiters or safety wrappers in the code to isolate user input from instructions.
- Capability inventory: The skill is capable of transmitting data to an external API endpoint using the Python
requestslibrary. - Sanitization: No evidence of programmatic input validation or sanitization was found in the code before the data is transmitted.
- [DATA_EXFILTRATION]: The skill performs network operations to
api.klingai.comusing the Pythonrequestslibrary to communicate with the Kling AI service. While this is the legitimate service provider, the domain is not included in the standard whitelist of allowed domains.
Audit Metadata