Skip to content

Novelty Check

Searches the literature to see whether a research idea has already been done.

Stars
18,482
Licence
MIT
Files
1 file

What it does

When you describe a research idea, this skill tells the agent to break the idea into core claims, search arXiv, Google Scholar, Semantic Scholar and recent conference papers for overlapping work, and read the abstracts of likely matches. It then produces a structured report with the closest prior papers, the specific difference (delta) from each, a novelty score, and a recommendation to proceed, proceed with caution, or abandon. It can also send the findings to a second model (via a Codex or similar reviewer tool) for an independent check.

Great for

  • Researchers deciding whether an idea is worth a pilot experiment before investing months
  • Students checking whether a project topic has already been published
  • Anyone who wants a quick literature sweep with a written summary of closest prior work

Skip it if

  • Final, authoritative novelty clearance: the skill's own rules say it is an early gate, and search coverage is limited to what the agent can find online, so missed papers are possible
  • Non-research tasks or general web research unrelated to academic novelty, where the verdict format does not apply
  • Situations where unpublished or confidential ideas must stay private, because the skill sends search queries to the web and may send a dossier of your method description to an external reviewer model

Try asking

  • “novelty check: I want to test whether using contrastive decoding for long-context summarization has been done before”
  • “有没有人做过 基于检索增强的代码补全 with uncertainty-aware reranking?”
  • “Check novelty of my idea: train a small verifier to filter agent tool calls before execution”
  • “Verify this research idea is novel before I start implementing it: ...”
Step-by-step guide: how to use Novelty CheckInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    novelty-check.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip novelty-check.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip novelty-check.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Describe the method, the problem it solves, and the mechanism in a few sentences so the agent can extract sharp claims.
  • Treat the verdict as a triage signal: verify the named closest papers yourself before committing time.
  • If your idea is sensitive, avoid sending full unpublished details through the reviewer route, or check which reviewer backend is configured first.

What's inside

  • SKILL.md

Redistributed unmodified from wanshuiyin/Auto-claude-code-research-in-sleep under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.