Generate ranked, testable research ideas from a user-provided direction.
npx skills add https://github.com/duypham2801/ThS_LLM --skill idea-creator-duypham2801
Please help me install this Agent Skill. Skill: idea-creator Source: https://github.com/duypham2801/ThS_LLM/tree/main/.claude/skills/idea-creator Command: npx skills add https://github.com/duypham2801/ThS_LLM --skill idea-creator-duypham2801
It helps you move from a broad research direction to a concrete, publishable set of ranked ideas with feasibility, novelty checks, and (when possible) pilot validation.
research-wiki/ exists, writes idea pages and builds edges/query packs to preserve learning over time.For example, if you want publishable directions for “radiology report grounded object detection,” this Skill will survey recent work, propose several grounded research hypotheses, validate likely novelty, and recommend 2–3 ideas to pilot with concrete minimum experiments.
Ask for ideas by saying: Generate and rank research ideas for chest X-ray phrase grounding with bounding box prediction, focusing on semi-supervised learning and pilot experiments within a limited GPU budget.
High-intent search queries and answers about installing and using this skill.
To generate publishable research ideas, you provide a broad research direction and the system surveys literature, identifies gaps, and produces ranked, testable proposals with feasibility and novelty checks. It outputs concrete hypotheses and minimum viable experiments.
Checking research hypothesis novelty involves scanning recent literature and local papers to map the field and identify structural gaps. The system performs deep novelty checks and critical review for top candidates to ensure likelihood of originality.
Yes, you can design pilot experiments with limited compute resources by specifying constraints like a limited GPU budget. The system applies pilot experiment constraints to recommend minimum viable experiments for the strongest research ideas.
Literature survey integration works by scanning local papers and searching recent literature to map the field. It identifies structural gaps through this landscape survey, which forms the foundation for generating and filtering testable research proposals.
No, you do not need a research-wiki directory to use the idea generation pipeline. Research-wiki integration is optional, but if the directory exists, the system writes idea pages and builds edges to preserve learning over time.
Automated experiment design for research ideation relies on LLM-based brainstorming and web search for recency, meaning output quality depends on available literature. It produces ranked proposals and minimum viable experiments, but cannot guarantee publication outcomes.
sjunglingMove and rename Obsidian notes while automatically updating internal wiki-links.
chenghaoYangGenerate and rank research ideas with hypotheses and experiments in an IDEA_REPORT.md.
LambenthanGenerate ranked empirical research ideas from the EmpiricalWiki knowledge graph.
hyperbolic-cGenerate structured, testable hypotheses with mechanistic explanations and falsifiable predictions from observations and literature syntheses.