Strategic research thinking agents for Claude Code — idea evaluation, project triage, and structured brainstorming to help you do research that matters.
This repository now also includes native Codex support.
Most AI writing tools help you write papers. This plugin helps you decide which papers to write.
Inspired by Nicholas Carlini's essay "How to Win a Best Paper Award" — which argues that great research starts with taste, strategic problem selection, honest self-evaluation, and knowing when to kill your darlings.
📢 Update (April 2026): Research Companion is now also a component of
researcher-pack— a more advanced, all-in-one Claude Code plugin for researchers that bundles strategic brainstorming, paper reading, literature tracking, academic writing, and project management into a single integrated workflow. This standalone repo is still actively maintained, and any updates to Research Companion will be synced between both repos — so you can use it here on its own, or get the full toolkit over atresearcher-pack. Feedback, issues, and PRs are very welcome in either place! 🙏
Researchers don't lack the ability to write papers. They lack a trusted colleague who will:
- Tell them an idea isn't worth 6 months of their life — before they invest those months
- Ask "who else is working on this and what's your unfair advantage?"
- Challenge them to state the key insight in one sentence (and refuse to move on until they can)
- Help them find the unexpected cross-field connection that makes a contribution truly novel
- Evaluate whether a struggling project should be continued, pivoted, or killed
This plugin provides that colleague.
| Agent | What it does |
|---|---|
| Idea Critic | Stress-tests research ideas along 7 dimensions: novelty, impact, timing, feasibility, competitive landscape, the nugget, and narrative potential. Returns a Pursue / Refine / Kill verdict. |
| Research Strategist | Project-level strategic thinking — triage (continue/pivot/kill), comparative advantage mapping, impact forecasting, opportunity cost analysis, and scooping risk assessment. |
| Brainstormer | Enhanced creative brainstormer with explicit focus on cross-field connections, "strategic ignorance" (challenging flawed assumptions the field follows uncritically), and the skeptical-reader test. |
| Skill | What it does |
|---|---|
/research-companion |
A structured multi-phase ideation session that orchestrates all three agents through: Seed → Diverge → Evaluate → Deepen → Frame → Decide. Includes Carlini's "conclusion-first test." |
8 research strategy principles organized into three categories (Problem Selection, Execution Strategy, Strategic Positioning) that guide the agents' evaluations.
This repository also includes a native Codex plugin manifest and interface metadata:
.codex-plugin/plugin.jsonagents/openai.yaml.agents/plugins/marketplace.json
The same core prompts, agents, and principles are shared across Claude Code and Codex, with the orchestration instructions written to degrade gracefully if delegated subagents are unavailable.
Codex supports plugin marketplaces and installation through the Plugin Directory. This repo now includes a repo-scoped marketplace file so the plugin can be installed after cloning, without manually creating marketplace JSON. Detailed Codex setup lives in docs/codex-installation.md.
claude plugin marketplace add https://github.com/andrehuang/research-companion
claude plugin install research-companion@andrehuang-research-companionSee docs/codex-installation.md for the full Codex installation guide, including:
- repo marketplace installation
- personal local marketplace installation
- official Codex plugin docs links
Just describe your idea and ask for evaluation:
I'm thinking about studying how LLM-generated code introduces subtle security
vulnerabilities that pass standard code review. Can you evaluate this idea?
The Idea Critic will evaluate across 7 dimensions and give you a verdict with the single most important question to resolve next.
I've been working on adversarial attacks against multimodal models for 3 months.
I have some results but they're incremental. Two other groups just posted preprints
in the same area. Should I continue?
The Research Strategist will assess your competitive position, impact potential, and opportunity cost, then recommend Continue / Pivot / Kill.
In Claude Code:
/research-companion I'm interested in the intersection of program synthesis
and scientific discovery
In Codex:
$research-companion I'm interested in the intersection of program synthesis
and scientific discovery
This launches a 6-phase guided session:
- Seed — Understand your problem space, interests, and what bugs you about the field
- Diverge — Generate ideas, alternative framings, and cross-field connections
- Evaluate — Stress-test the top 2-3 ideas with the Idea Critic
- Deepen — Check novelty, positioning, and competitive landscape
- Frame — Write the abstract and conclusion as if the paper is done (the conclusion-first test: if you can't write a compelling conclusion now, the idea isn't ready)
- Decide — Final assessment with next steps (starting with the single riskiest assumption to test first)
I work on differential privacy. What ideas from other fields
(cryptography, economics, ecology, etc.) could lead to novel approaches?
The Brainstormer is designed to bridge distant fields — the kind of connection that led to applying differential cryptanalysis to model stealing, or semi-supervised learning to poisoning attacks.
Here's my draft abstract. Play devil's advocate — what would a skeptical
Area Chair say?
The Idea Critic will identify the strongest counter-arguments, the weakest assumptions, and what a hostile but fair reviewer would target.
When the Idea Critic evaluates your research idea, it assesses:
| # | Dimension | Key Question |
|---|---|---|
| 1 | Novelty | If you don't do this, how long until someone else does? |
| 2 | Impact | Can you write a compelling conclusion right now, without doing the work? |
| 3 | Timing | Is the field ready for this? Too early? Already crowded? |
| 4 | Feasibility | What's the single riskiest assumption? Can you test it in a week? |
| 5 | Competitive Landscape | Who else is working on this? What's your unfair advantage? |
| 6 | The Nugget | Can you state the key insight in one sentence? |
| 7 | Narrative Potential | Can you tell a story that makes a skeptical reader care? |
These principles, derived from patterns in high-impact research, guide all agents in this plugin:
Problem Selection
- RS1: The Novelty Test — "If you don't do this, how many months until someone else does?"
- RS2: The Conclusion-First Test — "Can you write a compelling conclusion right now?"
- RS3: The Nugget Test — "Can you state the key insight in one sentence?"
Execution Strategy
- RS4: Fail Fast — "Start with the sub-problem most likely to kill the project"
- RS5: Kill Early — "A working project with low impact is worse than a killed project"
- RS6: Unreasonable Effort — "Strengthen 'sometimes' to 'usually' through additional work"
Strategic Positioning
- RS7: Comparative Advantage — "Research space is high-dimensional; find your unique corner"
- RS8: Timing Awareness — "Impact = skill x domain importance at this moment"
Evaluation results are now saved to disk so they persist across sessions:
- After each session, verdicts are written to
research-evaluations/YYYY-MM-DD-<topic>.md - On subsequent sessions, the system checks for prior evaluations of similar topics before starting fresh
- PARK'd ideas include "revisit conditions" — what would need to change to reconsider
- The Research Strategist now outputs a watch list (search terms, key researchers, venues to monitor) for competitive tracking
- The Idea Critic checks for prior evaluations to avoid re-evaluating killed ideas unless conditions have changed
This means your research thinking accumulates over time rather than starting from scratch each session.
- researcher-pack ⭐ — The recommended way to use Research Companion. An integrated Claude Code plugin for researchers that incorporates Research Companion as its strategic-thinking component, alongside paper reading, literature tracking, academic writing, and session management. If you use more than one of the tools below, install
researcher-packinstead of stitching them together by hand. - Academic Writing Agents — 12 agents for reviewing, auditing, drafting, and polishing academic papers. Research Companion handles the "what to write" question; Academic Writing Agents handles the "how to write it" question.
- Claude-Claw — OpenClaw-inspired enhancements for session continuity, task management, and memory architecture. Provides
/handofffor saving session state and/managerfor tracking research tasks.
Most researchers optimize for publication count. The best researchers optimize for impact — and papers follow naturally. This plugin is built around that distinction.
It won't help you produce more papers. It will help you produce better ones by being honest about which ideas are worth your finite time and by stress-testing your thinking before you commit months of effort.
The agents are deliberately opinionated and direct. They'd rather save you 3 months than spare your feelings. If an idea should be killed, they'll say so — and explain why.
MIT