The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references official Alibaba Wan model weights and code hosted on GitHub and Hugging Face. These are legitimate resources for the skill's stated purpose and originate from well-known official organizations.- [COMMAND_EXECUTION]: The provided Python client uses subprocess to execute ffprobe and ffmpeg for validating generated video files. The implementation includes significant security mitigations, such as using shell=False, enforcing execution timeouts, and limiting output buffer sizes to prevent resource exhaustion.- [DATA_EXFILTRATION]: Communication with Alibaba Cloud APIs is handled securely via environment variables for credentials. The script prevents unauthorized data egress by enforcing a whitelist of approved delivery hosts and performing IP address validation to block internal network access.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts which are then sent to an external generation API.
Ingestion points: User-provided prompts via the --prompt command-line argument in wan_hosted.py.
Boundary markers: Prompts are encapsulated within a JSON request body sent to the API.
Capability inventory: The script has network access (API calls), file system access (writing videos and logs), and command execution (ffprobe/ffmpeg).
Sanitization: The script performs length validation on prompts and rigorous validation of the resulting media files (signature, MIME, dimensions, and full decode checks) before final publication.

alibaba-wan-video