Guide · Agent skill
How to use the Research Idea Brainstorming agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The brainstorming-research-ideas skill is a prompt-based guide that gives an AI agent structured thinking methods for turning a vague research interest into a concrete proposal. It is aimed at academic researchers, PhD students, and teams preparing a brainstorming session. After reading this guide you will know what the skill does, how to install it, how to trigger it, and where it stops being useful.
Quick start
- 1Download the ZIP. Get the skill ZIP from the download page where you found this guide. It contains SKILL.md.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project.
- 3Or upload it in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the ZIP.
- 4Describe your research situation. Write a plain-language request that says what you are exploring, pivoting from, or stuck on.
- 5Work through one lens. Pick a lens such as problem-first vs. solution-first, and follow its workflow one step at a time.
- 6Check the self-check questions. Answer the listed questions honestly, for example whether a specific group needs the work.
- 7Verify novelty yourself. Search the literature for any candidate idea, since the skill does not confirm that ideas are unexplored.
What the skill makes the agent do
The skill provides structured ideation frameworks, which are step-by-step lenses for finding research directions. Its stated purpose is to help researchers move from vague curiosity to concrete, defensible research proposals. The SKILL.md describes ten lenses. The excerpt we reviewed covers five of them: problem-first versus solution-first thinking, the abstraction ladder, tension and contradiction hunting, cross-pollination from other fields, and the 'What Changed?' principle.
Each lens comes with a workflow and, in several cases, self-check questions. For example, the problem-first lens asks you to write your idea in one sentence, classify it as problem-first or solution-first, and then check whether the problem actually matters and who suffers from it.
- Problem-first vs. solution-first: checks whether the motivation is a real need or a tool looking for an application.
- Abstraction ladder: move a result up to a general principle, down to a constrained instance, or sideways into another field.
- Tension hunting: finds pairs of goals, such as performance and efficiency, that are often treated as trade-offs.
- Cross-pollination: borrows structural ideas from fields such as neuroscience, physics, economics, ecology, linguistics, and control theory.
- What Changed?: revisits older problems when compute, scale, data, regulation, or tooling has shifted.
Install the skill from the download page
The skill is distributed as a ZIP file from the page where you found this guide. Its only listed file is SKILL.md. There are two ways to install it, depending on how you use your agent.
For Claude Code, unzip the download into a skills folder. Use ~/.claude/skills/ to make the skill available to you everywhere, or a project's .claude/skills/ folder to make it available only inside that project.
In the Claude apps, you can upload the ZIP instead. Open Settings, go to Capabilities, then Skills, and add the skill there.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Trigger the skill with natural requests
You do not need a special command. The skill's description says to use it when you are exploring new problem spaces, pivoting between projects, or looking for novel angles on existing work. Describe your situation in plain language and the agent should pick up the skill.
Be specific about your situation. Mention what you are working on, what feels stuck, and whether you want to explore or evaluate a particular idea. The more context you give, the more useful the framework walkthrough will be.
- I'm starting a new research direction in retrieval. Help me explore the space with structured frameworks.
- I've hit a wall on my current project. Give me fresh angles on this existing work.
- I have a half-formed idea about model compression. Help me see whether it has real potential.
- I'm preparing a brainstorming session with my lab. Give me a structured agenda.
- I want to find trade-offs in my area that might be worth resolving.
What the skill does not do
The skill is a prompt-based guide only. It does not run code and does not fetch outside information. The SKILL.md lists no required dependencies, the skill does not run scripts, and it does not use the network.
It also does not check novelty. It generates ideas, but it does not search papers to confirm that an idea is unexplored. The skill tells you to use other skills for experimental design or methodology, and it redirects literature reviews to a separate literature-review skill, referenced as scientific-skills:literature-review.
- No code execution and no outside lookups
- No verified novelty checks against the literature
- Redirects experimental design to domain-specific skills
- Redirects literature reviews to scientific-skills:literature-review
Tips for getting useful output
Run one lens at a time when you are starting out. Combining lenses works too, but a single focused lens is easier to check. Treat the self-check questions as real questions. The SKILL.md asks, for instance, whether you can name a specific person or community who needs the work, and whether a claimed tension has been confirmed rather than assumed.
Verify the output yourself. The skill helps you structure thinking, but any candidate direction still needs a literature search and expert judgement before you commit time to it.
- Start with one sentence describing your current focus.
- Answer the self-check questions honestly before moving on.
- Use the output as an agenda for a group session.
- Check every proposed novelty claim against the literature yourself.
What people use it for
- A PhD student choosing a first project direction by testing ideas against the problem-first self-check.
- A lab lead preparing a brainstorming session who wants a structured agenda built from the lenses.
- A researcher stuck on an existing project who uses the abstraction ladder to move a specific result up to a general principle or down to a constrained case.
- A researcher evaluating a half-formed idea by listing competing goals and checking whether the trade-off is fundamental or an artifact of current methods.
- A researcher moving into a new field who borrows structural ideas from a neighbouring discipline using the cross-pollination workflow.
Frequently asked questions
How do I install the brainstorming-research-ideas skill?
Download the ZIP from the page where you found this guide and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.
Does the skill need anything else installed?
The SKILL.md lists no dependencies. The skill does not run scripts and does not use the network, so it needs no extra tools to run.
Is the brainstorming-research-ideas skill free?
The skill is released under the MIT licence, as stated on its page. Check the licence file in the download for the full terms.
Can this skill check whether my research idea is novel?
No. The skill generates ideas but does not search papers to confirm they are unexplored. You need to run your own literature search for novelty.
Can I use it for a literature review?
No. The skill redirects literature reviews to a separate skill, referenced as scientific-skills:literature-review.
How is it different from just asking an AI for research ideas?
The skill gives the agent specific frameworks with workflows and self-check questions, such as problem-first vs. solution-first classification and the abstraction ladder. A generic prompt does not guarantee that structure, although the output still needs your own judgement.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits PhD students and researchers who want a structured way to find or pressure-test a project direction, and teams that want a brainstorming agenda. It is not the right choice if you already have a defined research question and need experimental help, a literature review, or confirmed novelty checks, so pick a domain-specific or literature-review skill for those.
Download the skill