hegkfb tang

I independently design, prototype, and open-source production-ready AI system prompts and agent workflows. My work focuses on information boundary control, output evaluation, and conversational rules that prevent over-suggestion and factual overclaim.

In Open Fact Quick Diagnosis, I turned public-information verification into an executable AI workflow. I designed source reliability grading, claim levels, conflict comparison, mandatory justification, forbidden wording, and a self-check that blocks output until rules pass.

I built a runnable prototype using Python, Gradio, and an LLM API, then open-sourced it. I also created Xianying for self-exploration conversations and AmIStuck, a structured tool for evaluating agency, decision rights, and exit paths.

My cross-cultural study and work experiences inform my approach to communication and method-building. I bring Chinese-Japanese business translation experience and professional working proficiency in Japanese and English.