wip-grok_skill

Overview

This skill integrates xAI Grok capabilities to search the web and X, and to generate or edit images and short videos. It combines sensor functions (real-time web and X search) with actuators (text-to-image, image editing, and text-to-video) so an agent can both gather current information and produce media assets. The skill is focused on practical, real-time tasks rather than long-term archival or heavy numerical computation.

How this skill works

Search functions return content, citations, and raw responses for web and X queries, with options to restrict domains or handles and enable image/video understanding. Generation functions produce images or short videos from prompts, accept seed images for edits or animation, and return temporary URLs or base64 data. Video generation is asynchronous and uses a request_id you poll or wait on until the job completes.

When to use it

• Get current news, documentation, or real-time facts from websites.
• Monitor sentiment, trends, or reactions on X/Twitter around a topic or user.
• Create concept images from text prompts or generate multiple visual variations.
• Edit existing images with natural-language instructions or combine up to three source images.
• Produce short (1–15s) videos or animate still images for prototypes or social clips.

Best practices

• Specify allowed/excluded domains or X handles to focus results and reduce noise.
• Use enable_image_understanding or enable_video_understanding when media context matters.
• Download image and video URLs immediately—they are temporary and expire.
• Limit n (number of outputs) to control cost; track cost per image/video in workflows.
• For video, use wait_for_video or poll_video to handle asynchronous completion and retries.

Example use cases

• Compile a brief of latest press and social reactions about a product launch using search_web + search_x.
• Generate multiple logo or concept image variants for stakeholder review with generate_image.
• Edit product photos to change color, background, or combine elements using edit_image.
• Create a 10-second promotional animation from a brand image with generate_video and wait_for_video.
• Monitor a specific X handle for breaking commentary and summarize sentiment in real time.