Developer turns China’s “OpenClaw (小龙虾)” fad into six practical, open‑source skills

Lead

OpenClaw (小龙虾) — an agent/skill layer that has exploded across Chinese social circles — is mostly being used as a chat toy, but one developer has repurposed it into six genuinely useful automation skills and published the full source code on GitHub. The angle is simple: instead of using the agent to flirt or “auto‑trade” (claims that are reportedly exaggerated and often blocked by platform moderation), use it to save hours of manual work. And yes, the repos are public.

What the skills do

The projects include a photo watermark generator (multi‑brand EXIF and model stamping) for batch processing images (https://github.com/howillli/photo-watermark); a tech‑news aggregator that scrapes, summarizes and renders daily headlines into long images for quick consumption (https://github.com/howillli/tech-news-aggregator); a video cover generator that makes horizontal and vertical covers from a script (https://github.com/howillli/video-cover-generator); an image‑prompt reverse tool that produces reproducible prompts for Midjourney, DALL·E and Stable Diffusion (https://github.com/howillli/image-prompt-reverse); and a person resume search that compiles public web data into a visual HTML CV (https://github.com/howillli/person-resume-searche). The author also describes a daily video‑tracking workflow to study hot‑video trends across platforms such as Bilibili (哔哩哔哩).

Why Western readers should care

If you’re unfamiliar with China’s rapidly iterating AI tooling scene: these agent layers are where many users first meet large‑model workflows. But Chinese platforms enforce strict content and account rules, so the developer optimizes for token efficiency and deterministic rendering to avoid generation errors and moderation hits. It has been reported that AI‑generated images sometimes outperform original photos in engagement, a claim the author says he observed firsthand — but platform rules and latency (for example, token generation speed versus market movement) make automated trading both risky and impractical.

Code and caution

All six projects are released with source on GitHub for anyone to inspect and adapt. Want to automate your daily digest or batch‑watermark a photo library? These tools make it possible — but proceed with care: privacy, copyright and platform‑policy risks remain real in China’s tightly policed social ecosystem, and broader geopolitical factors (such as export controls on advanced compute) continue to shape what domestic AI tools can realistically do.