Rank candidate research papers from anchor, topic, wiki, or venue inputs.
npx skills add https://github.com/skyllwt/AutoSci --skill discover-skyllwt
Please help me install this Agent Skill. Skill: /discover Source: https://github.com/skyllwt/AutoSci/tree/main/.claude/skills/discover Command: npx skills add https://github.com/skyllwt/AutoSci --skill discover-skyllwt
It helps you quickly find the next papers to read from your existing context by producing a ranked, rationale-backed shortlist instead of leaving you to guess what’s relevant.
Run the discovery for a topic to get a ranked shortlist: "/discover --topic diffusion model fine-tuning --limit 10".
High-intent search queries and answers about installing and using this skill.
You can get ranked research paper recommendations by running a discovery command that generates a proposal-only shortlist. The tool logs a discover checkpoint and optional wiki log without downloading papers or mutating existing wiki content.
Find related papers from a specific venue or year by providing conference inputs to discovery. The tool compares candidates against your existing wiki content, deduplicates entries, and returns a ranked shortlist of relevant works.
Yes, you can generate a reading shortlist using anchor-driven discovery from specific seed papers. You can also provide negative seeds to refine recommendations, ensuring candidates match your current research context.
Discover relevant papers by providing a free-form topic string to the discovery tool. It produces a rationale-backed, ranked shortlist of candidate papers without ingesting any content into your wiki.
Yes, the discovery tool performs wiki deduplication to ensure proposed candidates do not overlap with existing wiki paper pages. It maintains safety-first behavior by writing proposals and logs without mutating wiki content.
The proposal-only discovery workflow never ingests content, downloads papers, or mutates wiki content. It is limited to producing a ranked shortlist and logging the run, ensuring you manually review candidates before any ingestion occurs.
duany049Rank candidate research papers from anchor, topic, wiki, or venue/year seeds.
Devsoul2026Consolidate diverse sources into an interlinked Markdown wiki with provenance.
ashiqcodeleafSearch arXiv.org by keyword, author, category, or ID and fetch paper metadata.
kamilseghrouchniSearch the wiki with a request.json filter and produce candidates.json and discover_report.md.
