Generate images and videos through AI-video-agent. Supports image create, image remix, video create, and video animate.

This package appears to be a straightforward client/CLI for an AI-video-agent backend. Before installing or running it: (1) Only set AI_VIDEO_AGENT_BASE_URL to a trusted service — local files you pass (reference images, first frames) will be uploaded to that service. (2) Treat AI_VIDEO_AGENT_TOKEN as a secret and avoid pasting it into logs; the bundled login helper prints a token to stdout, which could be captured in terminal history or CI logs. (3) Result URLs returned by the service will be downloaded and saved locally (./output). (4) There is a minor metadata mismatch: the registry notes 'instruction-only', but the package includes a full Python client and scripts — this just means the skill contains code (no installer), not that it reaches out to unexpected endpoints. If you do not trust the AI-video-agent server you configure, do not provide any sensitive local files or tokens.

Type: OpenClaw Skill Name: canvas-claw Version: 0.1.0 The canvas-claw skill bundle is a legitimate tool designed to interface with an AI video and image generation service. The code is well-structured, follows standard Python practices, and includes comprehensive documentation and tests. While it possesses the capability to upload local files to a user-configured API endpoint (implemented in `canvas_claw/media.py` and used by `generate_image.py` and `generate_video.py`), this behavior is explicitly documented as necessary for features like image remixing and video animation. No evidence of malicious intent, data exfiltration to unauthorized domains, or prompt injection attacks was found.