Search and screen arXiv papers for a target field and time window, filter by institution allowlists, then rank open-source papers by GitHub stars and non-open-source papers by citation count. Use when Codex needs to retrieve recent arXiv papers for a field such as spatial intelligence, world model, or agents, and narrow them to papers from top universities, major AI/internet companies, or field-specific leading institutions.
Build a weekly Hugging Face papers digest for a specified week id (`wid`) and a set of themes. Use when Codex needs to collect papers from a weekly Hugging Face papers page, filter them into primary themes, sort by upvotes, generate a themed summary table with key images when available, summarize cross-paper common ideas for each theme, and write one standalone markdown analysis for the highest-upvoted paper in each theme.
Publish a Xiaohongshu image note from a normal Safari window that is already logged into the creator platform. Use when Codex already has the final title, body, and image assets, and needs to post one 图文 note through the validated Safari creator flow.
Prepare Xiaohongshu post copy and image assets for either overview posts or single-paper analysis posts. Use when Codex needs to draft the post, choose titles, prepare markdown, and collect the right cover/figure images before any Safari publishing step.