StatsClaw plugin
StatsClaw assembles nine Claude Code agents to build statistical packages from a paper or formulas with isolated pipelines.
- 101 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Agents write code and push commits and PRs to GitHub on your behalf, which needs write access.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/statsclawDetects 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.
- Agents write code and push commits and PRs to GitHub on your behalf, which needs write access.
This is third-party code. Review the repository files before installing.
What it does
The framework for Claude Code installs an orchestration policy in `CLAUDE.md`, nine agent definitions, shared skills, language profiles and templates. A leader dispatches work, a planner reads a paper or formulas and writes specifications, a builder writes code from `spec.md` only, a tester validates from its own `test-spec.md`, and a simulator runs Monte Carlo studies. A reviewer cross-checks every pipeline and issues a verdict, and a shipper handles the commit and PR. The code, test and simulation pipelines are isolated and never see each other's specs. If they converge independently, confidence in correctness is higher. Profiles cover R, Python, Julia, TypeScript, Stata, Go, Rust, C and C++.
Who it is for. Researchers and statistical package authors who turn methods from papers into R, Python or Julia code.
Good fit when
- A method from a paper must become a package with tests.
- A bug is fixed or a feature is added to an existing package.
- A package is translated between languages, for example R to Python.
- Monte Carlo studies and documentation are needed alongside the code.
Not a fit when
- You need one short edit instead of nine roles.
- The repository cannot be given to an agent with write and push access to GitHub.
- The task has nothing to do with statistics and scientific packages.
Example request
Implement the method from the attached paper as an R package, write tests from a separate specification and prepare a Monte Carlo simulation.Limitations
It needs Claude Code, write access to the target GitHub repository and a workspace repository for artifacts, created automatically if needed. Agent Teams is enabled through `.claude/settings.json`. Nine agents spend noticeably more tokens than a single session. The method relies on independent convergence of pipelines, which is not a proof of correctness. For complex domain bugs the README itself calls for human judgment, and the researcher must always review the result.
How to disable. Uninstall the plugin with `/plugin uninstall` or remove `CLAUDE.md`, `agents/`, `skills/` and `profiles/` from the project and turn off Agent Teams in `.claude/settings.json`.
FAQ
Why are the pipelines isolated?
The builder never sees the test spec and the tester never sees the code spec. If results converge independently, it is a check from opposite sides.
Which languages are supported?
Profiles exist for R, Python, Julia, TypeScript, Stata, Go, Rust, C and C++.
What does a task produce?
Planner specifications, code, tests, simulations when needed, documentation and an audit trail. The shipper opens the commit and PR.
Is a human needed in the process?
Yes. The framework asks questions on ambiguity and the researcher validates the mathematics.
What can I ask in one sentence?
For example, to work on a repository and resolve its issues. The leader detects the language and workflow itself.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code plugin for modern R: eight skills, commands, agents and hooks for tidyverse, rlang, S7, brms and testing
MCP server that connects Claude Code and Codex to an open R or Python session in RStudio and Positron, where your data lives
A Claude Code agent for protein design: target research, antibody and nanobody generation, screening and candidate ranking.
26 skills and four reviewer agents for science in Claude Code: figure and statistics review, literature search, work handoff.
A set of commands, skills and subagents for Claude Code that runs data analysis like a careful team analyst.
43 skills for Claude Code along the research cycle: surveys, experiments, topic models, reporting and manuscript review
README in short
The README describes the framework idea, a table of nine agents, a multi-pipeline architecture diagram and key properties. A quick start with prerequisites, the list of installed files and working modes follow.