shidi-skill

Daily research is full of repetitive, grunt-work tasks—from mining literature and designing experiments to plotting figures and deep-reading papers—that drain mental energy after the initial spark of an idea. shidi-skill is built to tackle this very problem. It isn't a mere prompt template or chatbot, but a complete research workflow embedded within your AI IDE. You hand it a research idea, and it autonomously executes tasks like literature surveys, experiment design, data analysis, and figure generation, ultimately delivering structured reports, verifiable data files, and a cross-verification brief. This transforms abstract 'chat answers' into tangible, deliverable outcomes.

Its core capabilities span the research lifecycle: literature research launched with three parameters (supporting 9-angle diffusion, six-dimensional scoring, and batch PDF download), experiment design with explicit vulnerability checklists, publication-quality figure generation from real data, and structured deep reading of papers. The process is fundamentally anti-hallucination, enforced through triple deduplication (DOI/arXiv/URL), transparent reading-depth labels, and a mandatory cross-verification step. All data and files are generated locally, ensuring no platform or model lock-in, so you retain full ownership of your research assets.

If you are a researcher in AI for Science, a STEM graduate student, or anyone using agentic AI to augment your workflow, shidi acts as your humble yet efficient 'digital junior lab mate.' It never makes decisions for you—angle selection, parameter confirmation, and trade-offs remain your call—while it handles the legwork and delivers files. When you say, 'Shidi, help me with [topic],' it gets to work, asks only what it must, and hands you the results, freeing you to focus on higher-level scientific judgment and innovation.