What problem does it solve? Publishing research artifacts (code, data, experiment logs) in a citable, archived form is tedious: you must package the bundle, upload it to a dataset repository, register a DOI, and record the publication metadata. This Skill automates that final publication step of the auto-research workflow. ## Core Features & Use Cases - HuggingFace Hub Datasets upload: Creates a dataset repo and uploads the exported bundle via the huggingface_hub API, with private/draft mode support. - Zenodo archiving with DOI: Creates a deposition, uploads the bundle, sets metadata (title, creators, keywords), and optionally publishes to obtain a DOI, with sandbox mode for testing. - Publication record generation: Writes PUBLICATION.md with URLs, DOI, git SHA, and a BibTeX citation, and updates STATE.json with publication metadata. - Use Case: After completing Phase 8 of an LLM research project, publish the exported bundle to both HF Hub and Zenodo so reviewers receive a DOI-linked, reproducible artifact. ## Quick Start Use the research.publish skill to upload the latest export bundle for my project to HuggingFace Hub and Zenodo in draft mode.