ClawAgent Content Creation Assistant

Problem

Content workflows for e-commerce and social media often stall across asset preparation, try-on display, posters, notes, and video publishing. ClawAgent groups several content-generation tools into one callable flow: user info, AI try-on for clothes and shoes, model generation, product scene images, posters, Xiaohongshu notes, and tasks such as voiceover, singing, mixed editing, video cloning, link parsing, and video analysis. It is useful when you need to connect local assets, generated results, task status, and sharing steps into a repeatable pipeline.

How It Works

• Invocation: centered on /agent_mcp.cjs; parameters can be piped with echo '' |, and the Schema should be checked when arguments are unclear.
• Core tools: model_try_clothes, shoes_dressing, model_produce, business_product_scene, business_poster, ice_design_image_xhs, ice_voice_video, image_song, ai_mixed_script, ai_clone_video, get_raw_url, and analyze_video.
• File input: APIs expect public URLs; local files must be uploaded first and returned as FILE_URL; guessed paths or sample URLs are not acceptable.
• Async jobs: poll with get_job_status, using the documented sleep values for image, mixed-editing, and voiceover tasks, up to 20 attempts.
• Errors: when code != 0, read msg; handle missing_token, 403, and insufficient compute by telling the user what happened.

Boundaries

This skill depends on a token, compute credits, and external generation services. Some tools require accessible input links. A video URL may appear while the job is still processing, so status should be the source of truth before showing results.