How to use the Creative Thinking for Research agent skill

What the skill makes the agent do

The skill packages eight frameworks from cognitive science, each targeting a distinct cognitive operation: combining, reformulating, analogizing, constraining, inverting, abstracting, exploring boundaries and holding contradictions. Examples include bisociation (connecting two previously unrelated frames of reference), problem reformulation (re-representing the problem rather than solving it as stated) and structure-mapping analogies (matching deep relational structure across domains).

Instead of offering open-ended brainstorming, the agent works through these methods step by step. The skill describes the output as raw insight: ideas you then evaluate and structure further.

Install the skill from the ZIP or the Claude apps

Download the ZIP of the skill from this page and unzip it. Place the resulting folder in ~/.claude/skills/ for personal use, or in a project's .claude/skills/ folder to share it with one project.

Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• App install: Settings → Capabilities → Skills

Trigger the skill with natural requests

The skill's description says to use it when seeking genuinely novel research directions. You do not need a special command: describe what you want in plain language that matches that purpose, such as asking for novel directions that combine ideas from two fields, or asking for a way out of a rut within one subfield.

The skill is for ideation only. Phrasing that asks for a literature survey or for executing a defined problem falls outside its scope.

• Ask for cross-field directions: find novel research directions that combine ideas from two fields you know
• Ask for a reframing: help me re-represent this problem; what assumptions am I making?
• Ask for analogies: find structural analogies from other domains for this bottleneck

Use it for the right tasks

The skill suits PhD students or researchers stuck within one subfield, research leads preparing an ideation session or retreat, and developers exploring non-incremental approaches to a hard algorithmic or ML problem.

It is not the right tool for structured project brainstorming (the skill says to use the separate brainstorming-research-ideas skill for that), for executing a well-defined research problem, or for literature surveys (the skill points to a separate literature-review skill). The skill also notes that ideas generated this way still need checking against existing work.

Requirements, scripts and network use

The skill lists no requirements and no dependencies. Its only file is SKILL.md. It does not run scripts and does not use the network, so there are no setup steps beyond installing the folder.

Because it runs no code and makes no network calls, the main safety consideration is the quality of the output: verify every idea against existing literature before acting on it.

• Requires: nothing listed
• Runs scripts: no
• Uses network: no
• Files: SKILL.md

Tips for better results

Pair this skill with the brainstorm skill: use creative-thinking to generate raw insight, then brainstorm to structure and evaluate it. The skill's own workflow suggests this division of labour.

Describe your problem in relational language, stripping domain-specific nouns, before asking for analogies. The skill's example turns a transformer attention problem into a description of selective aggregation over a set, which opens more structural matches.

• Give two domains you know at least passingly when asking for bisociation
• Ask the agent to test each candidate for structural depth, not just surface similarity
• Check every promising idea against existing work before investing time