Science Superpowers plugin
16 research methodology skills for agents: from a testable question and pre-registered hypotheses to reproducible analysis
- 364 ★ on the repository
- Last commit
Extended access
Why this level
- The session hook runs Bash at the start of every session and injects instructions into the agent
- Following the skills the agent runs computations and creates working folders and files in the project
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/science-superpowersDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The session hook runs Bash at the start of every session and injects instructions into the agent
- Following the skills the agent runs computations and creates working folders and files in the project
This is third-party code. Review the repository files before installing.
What it does
Science Superpowers is a plugin that carries the Superpowers approach over from software development to computational science. When the agent sees you are about to investigate something, it first turns the interest into a testable question and only then touches your data. It then grounds the work in prior literature and standard methods, designs the analysis and pre-registers hypotheses, predictions and decision rules before it sees any outcome. That discipline replaces test-driven development here and guards against p-hacking and HARKing. The agent then runs the plan in a reproducible workspace: a pinned environment, fixed seeds and immutable raw data. When a result looks wrong it looks for the cause and keeps inconvenient data. Before claiming findings it reproduces the evidence afresh and then sends the work to a skeptical reviewer. The `prereg.sh` script freezes and audits predictions: from git history alone it shows that predictions preceded outputs and were not edited after the freeze. The set has 16 skills in all, and they trigger automatically.
Who it is for. For researchers, analysts and scientists who run computational experiments and data analysis with Claude Code, Cursor, Codex or Gemini CLI agents and want reproducible, checkable results.
Good fit when
- You start research from a loose interest and want a testable question with hypotheses out of it
- Before looking at results you need to lock hypotheses and decision rules and later prove it with git history
- A result looks implausible and you need to find the cause without discarding inconvenient data
- Before reporting you need to rerun the calculations and check assumptions and robustness
- The work needs splitting into steps handed to subagents with a fresh context
Not a fit when
- You write ordinary software: the original Superpowers exists for development
- You need a ready science package: the set installs no Python, R, solvers or scientific libraries
- You need a quick exploratory run with no process: a separate opt-in feasibility mode exists for that
- You have no Bash and no Git
Example request
I want to find out whether solver step size affects simulation stability. Help me frame the question and pre-register hypotheses before any runLimitations
- Installation is separate for each agent you use. A complete setup means discovery of the skills plus loading of the using-science-superpowers bootstrap skill.
- The Cursor and Claude Code hooks and the `prereg.sh` script need Bash and standard shell utilities, and `prereg.sh` also needs Git. Without Bash on Windows the hook exits and no bootstrap is injected.
- The set installs no Python, R, solvers or scientific packages: pin the environment in your own project.
- The plugin version in the manifest is 0.1.0. The author states the MIT license, but GitHub does not recognize it.
- Gemini CLI takes the set as an extension, and OpenCode and Pi have separate adapters.
How to disable. In Claude Code run `/plugin uninstall science-superpowers@science-superpowers-dev` and start a new session. For Cursor remove the link from ~/.cursor/plugins/local/. For other agents remove the package or extension the way you installed it. After that the session hook stops injecting instructions.
FAQ
How do I install Science Superpowers in Claude Code?
Inside Claude Code run `/plugin marketplace add K-Dense-AI/science-superpowers`, then `/plugin install science-superpowers@science-superpowers-dev`. The SessionStart hook loads the bootstrap, so open a new session.
What is pre-registration and why does an agent need it?
It means locking hypotheses, directional predictions and decision rules before any outcome is seen. The authors call the preregistering-analysis skill the set's Iron Law: it separates confirmatory from exploratory work and guards against p-hacking and HARKing.
How can I show predictions were not edited afterwards?
The `prereg.sh` script in the pre-registration skill freezes predictions and audits them. Besides Git it needs a POSIX shell and standard utilities such as awk, sed and diff.
What if a result looks strange?
The investigating-anomalous-results skill runs a four-phase root-cause search before any adjustment. Before conclusions, verifying-results-before-claiming asks you to rerun calculations and check assumptions.
Do I have to call the skills myself?
Usually not: the bootstrap loads at session start and the others trigger automatically when the agent sees a research task. If the hook is missing, invoke using-science-superpowers by hand at the start of a session.
Discussion
No questions or reviews yet.
Related
Plugins166 skills for scientific work: bioinformatics, cheminformatics, clinical data, geospatial analysis and 100+ databases
A skill and Python package for hyperparameter tuning: the agent reads the code, proposes the next set of values, and your own run measures the result
Auto-Empirical Research Skills
Auto-Empirical Research Skills (AERS)
A plugin marketplace for empirical social science research: causal inference, econometrics and manuscript prep for top journals
A Google DeepMind skill collection for science tasks: access to biology and chemistry databases, literature search and structural biology
A Claude Code plugin and MCP server: over 1000 scientific tools for biology, chemistry and medicine, data-check commands and routing skills
A library of agent skills for medical research: literature search and appraisal, study design, data analysis and academic writing
README in short
The set's workflow has ten steps, from the question to review, and the authors explain its philosophy separately with links to their blog. Each supported agent has its own installation guide.