The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and image file paths for transmission to the Kling AI API, which represents an indirect prompt injection surface.\n
Ingestion points: User-defined prompts and reference image paths ingested via the styled_generation and style_from_reference functions.\n
Boundary markers: The skill concatenates style modifiers with user prompts in the _build_styled_prompt method without using specific delimiters or safety instructions to isolate untrusted input.\n
Capability inventory: The skill utilizes requests.post for external API communication and open() for reading local image files.\n
Sanitization: No explicit sanitization or validation of user-provided prompt strings or file paths is implemented in the provided Python scripts.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill retrieves API access keys from environment variables (KLING_ACCESS_KEY, KLING_SECRET_KEY, KLINGAI_API_KEY), which is a secure method for handling sensitive credentials.\n- [DYNAMIC_EXECUTION]: The skill uses the base64 and jwt libraries for standard data transformation and authentication required by the Kling AI service.

klingai-style-transfer