Thesis AI Assistant

Problem

Thesis writing often stalls because the workflow is fragmented: topic scoping, literature search, reading notes, research design, data analysis, figures, citation formatting, typesetting, and defense materials live in different tools and chats. The assistant splits the thesis into 10 handoff-ready stages: startup diagnosis, topic selection, search, reading, knowledge base, research design, data analysis, figures, writing, and revision/defense. Each stage has a lead agent, supporting agents, and quality gates, turning abstract capability categories into concrete inputs, actions, outputs, and handoff rules.

How It Works

It uses an orchestrated skill: it receives the topic, current stage, and existing materials, determines which stage applies, loads the relevant agent role card, and executes the stage protocol. Artifacts are stored in a local thesis workspace, acting as shared memory across stages. Key flows include:
- Diagnosis: clarify degree level, discipline/language, and current state to choose the entry point.
- Literature loop: connect searching, structured paper breakdown, knowledge-base notes, method matching, and citation labeling.
- Data and figures: keep data sources, method notes, and figure requirements aligned before writing.
- Writing and review: use literature-driven writing, then move to pre-review, similarity checks, AI-rate checks, typesetting, and presentation materials.

If companion skills are not installed, the workflow can degrade to manual or expert-only modes, but formal database retrieval, publication-ready figures, typesetting, and PPT generation rely on external capabilities. It does not replace academic databases, nor does it fabricate references, DOIs, authors, or original viewpoints; factual claims need sources and user verification.