Plan a study
Use the experimental design guidance to define the question, variables, comparison groups, and sources of bias.
Scientific Agent Skills
K-Dense AI’s Scientific Agent Skills is an installable research collection for supported coding agents. Start with your research question and expected deliverable, then select a Skill whose inputs and environment fit the task.
Select One Scientific Agent Skill for Your Research Task
Define the research stage, available materials, and one artifact before selecting a workflow.
Provide the study stage, approved inputs, preferred host, expected output, and any restrictions on network or credentials.
Copy the request into a new Floatboat conversation and provide the required material.
Download FloatboatYour request for Floatboat
Preview and copy the request. It is not sent automatically from this page.
Reference: Scientific Agent Skills https://github.com/K-Dense-AI/scientific-agent-skills Read the original documentation and check the current environment and dependencies. Propose any new installation or tool connection before proceeding. My task: Read the official Scientific Agent Skills catalog. For my study-design brief, pick one suitable Skill and check its SKILL.md for license, dependencies, and output. Propose a local-only workflow and get my approval before installing or running anything. Expected deliverable: A Skill selection with source, license, prerequisites, and a bounded workflow for a reviewer to approve.
About Scientific Agent Skills
Scientific Agent Skills organizes research procedures into individual manifests and supporting files. K-Dense AI’s catalog groups workflows by task, including experimental design, data analysis, visualization, and scientific writing.
The upstream README documents installation for hosts including Codex, Cursor, and Claude Code. Choose a small subset and read each SKILL.md before use. Some Skills provide instructions alone, while others include scripts or depend on local libraries, reference data, or external services.
This overview helps select a workflow. For inspecting an approved local CSV or JSON file, use the separate Exploratory Data Analysis guide. For study planning or writing, identify the relevant research inputs and keep the human review step visible.
Choose a Skill by Its Research Output
Decide whether the next artifact is a study design, an analysis record, or a draft. Match the Skill’s prerequisites to that output before installing dependencies or sharing data.
Select an analysis
Use statistical analysis guidance to check assumptions and specify the method, effect size, and uncertainty.
Reconcile a manuscript
Use scientific writing guidance to align methods, reported results, figures, and citations with the supplied evidence.
Scientific Agent Skills Use Cases Across Research Stages
Choose a different Skill when the research decision changes. A protocol, an analysis record, a figure, and a manuscript revision each need their own inputs and checks.
Prepare an experimental design for review
Give the agent a research question, candidate outcomes, and operational constraints. Ask it to draft a design brief covering controls, randomization where relevant, possible confounders, and decisions the research team still needs to make.
Document a statistical comparison
Provide approved study data, group definitions, and the comparison you want to make. Use the statistical analysis guidance to choose a method, record assumption checks, and report an effect estimate with its uncertainty.
Build a figure from checked results
Supply a verified results table, units, and the purpose of the figure. Use a visualization Skill to create the figure and a reproducible script, then check axis labels, transformations, and values against the source table.
Review a manuscript draft for consistency
Provide an approved draft and the tables or outputs supporting it. Ask the writing Skill to identify mismatches between methods, text, and reported values. Keep unsupported citations or results marked for human follow-up.
Requirements Before Installing a Research Skill
Start with the selected Skill’s manifest and an approved workspace. Check execution tools and data access separately before choosing a local or networked workflow.
Supported Agent Skills host
The repository documents the npx skills installer for Codex, Cursor, and Claude Code. For a pinned individual install, see the upstream GitHub CLI instructions. Confirm the Skill is visible in the host.
Task-specific environment
Check required packages, scripts, reference files, credentials, and network access in the individual manifest.
Approved inputs and license
Use only data and services you may access. Check the Skill’s terms and review the planned output before sharing it.
Related Software Projects
Compare the workflows, operating requirements, and source licenses of these projects.
Scientific Agent Skills FAQ
Do I need to install the whole collection?
No. The upstream guide supports installing a selected Skill and recommends reviewing each manifest before adding it.
Does installation include API credentials?
Credentials, accounts, and network permissions depend on the selected workflow and need separate setup.
Can every Skill run offline?
Some workflows can run locally once their dependencies and data are available. Database or hosted-service workflows may require network access.
Does the repository’s MIT license apply to every Skill?
The main repository declares MIT. Check the license field of the specific Skill and any external tools or data it uses.
Where should I start with an unfamiliar CSV?
Start with the dedicated Exploratory Data Analysis guide and use an approved local file. The suite overview covers choosing the next workflow.
Can the agent approve the study or verify its findings?
Use the Skill to draft or run a bounded workflow. The research team reviews assumptions, evidence, and conclusions before relying on the output.