The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates through the execution of Python scripts scripts/aigc_videogen_multi.py and scripts/onboarding.py to manage video generation tasks and user account states.
[EXTERNAL_DOWNLOADS]: The tool downloads generated media assets from remote vendor-controlled URLs to the local session workspace.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data that could be exploited to influence agent behavior.
Ingestion points: The imageList and prompt fields provided by the user and processed in scripts/aigc_videogen_multi.py.
Boundary markers: Data is passed to scripts in a structured JSON format, providing basic structural boundaries, but there are no explicit delimiters within the free-text prompt field to distinguish instructions from data.
Capability inventory: The skill can perform network operations targeting vendor APIs and write files to the session's media/ and data/ directories.
Sanitization: Input parameters are serialized as JSON to prevent shell injection, but the content of the text prompt is not sanitized for potential embedded instructions.

linkfox-aigc-videogen-multi