What problem does it solve? Turning a raw research idea or proposal into a complete, compiled academic paper with real citations, editable vector figures, and verified numbers normally requires weeks of manual writing, citation hunting, and LaTeX debugging. This Skill orchestrates that entire pipeline automatically while enforcing strict integrity rules against fabricated results. ## Core Features & Use Cases - Input routing and staged pipeline: Classifies the dropped input (bare idea, structured proposal, or proposal with real results) and runs plan, cite, write, refine, review, figure, and LaTeX assembly stages, producing a compiled main.pdf. - Citation and integrity enforcement: Builds a refs.bib of 40+ real, fully-specified references via web search, and machine-checks that every reported number traces to the user's actual data in data-aware mode. - Editable vector figures and templates: Routes figures to matplotlib (results plots) or an image model with vision critique, then vectorizes them to editable PDFs; ships template-agnostic specs for Traitement du Signal and NeurIPS styles. - Use Case: Drop a one-line research idea and receive a compiled, citation-complete journal-format PDF with adversarially reviewed text and editable vector figures, ready for further experiment-driven revision. ## Quick Start Ask the assistant to turn your research proposal or idea into a complete compiled paper using the ts-paper skill, optionally naming a template such as ts_iieta or neurips.
