The Agent Skills Directory

[SAFE]: The skill manages API credentials through environment variables (e.g., DASHSCOPE_API_KEY), ensuring keys are not hardcoded or exposed in command-line arguments.
[EXTERNAL_DOWNLOADS]: The skill downloads generated video files and interacts with official AI provider APIs using the requests library. These operations are restricted to the primary functionality of video generation and target well-known cloud providers (Alibaba Cloud, Tencent, MiniMax, and Volcengine).
[DATA_EXFILTRATION]: No unauthorized network activity or data leakage was detected. File system operations are limited to reading input images and writing output videos and a local task cache at ~/.cache/videogen.
[PROMPT_INJECTION]: The skill ingests user prompts and images for video generation. While it lacks explicit prompt delimiters for the LLM-based backends, it enforces validation on media inputs. \n
Ingestion points: User input provided via the CLI prompt argument and --image/--ref flags in scripts/generate_video.py. \n
Boundary markers: None. \n
Capability inventory: Requests network operations and local file writes. \n
Sanitization: Image size limit (20MB) and extension checks.

videogencn