r-skills

r-skills

Claude Code plugin providing skills for R programming. Complementary to Posit skills.

Installation

Via Marketplace (recommended)

After pushing to GitHub:

/plugin marketplace add jsperger/llm-r-skills
/plugin install r-skills@r-skills

Local Development

claude --plugin-dir /path/to/r-skills

Dependencies

For full functionality, the following are required:

• languageserver - R package for LSP support
• jq - JSON processor for hooks
• air - R code formatter

Using the plugin effectively

• Formatting is automatic. A hook runs air format after every file edit. Never ask Claude to fix style or formatting — it is already done, and the language server is configured not to compete with it.
• Lint diagnostics are correctness-only by default. In a project with no .lintr/.lintr.R, the language server lints with the plugin's slim profile (config/agent.lintr): object-usage errors, missing packages, == NA comparisons, and similar real bugs — no line-length or spacing noise burning context. If your project ships its own .lintr, it always wins, and Claude sees exactly the lints your CI sees.
• Something not working? Run /r-lsp-diagnose. It checks the toolchain end to end, including the languageserver/lintr version combination that silently disables .lintr files.

Skills

designing-tidy-r-functions

Guidelines for designing user-friendly R function APIs, covering naming conventions, argument ordering, and output stability.

ggplot2

ggplot2 4.0+ features including S7 migration, theme defaults, and new scale/position aesthetics.

hardhat

Infrastructure for building tidymodels-compatible modeling packages using mold() and forge().

metaprogramming

Techniques for manipulating R expressions using rlang: defuse-and-inject pattern, quosures, and symbol construction.

r-languageserver

Using the R language server to navigate code: find definitions and references, understand call hierarchies, and run impact analysis before refactoring. Helps Claude discover LSP functionality the plugin alone did not surface.

rlang-conditions

Error handling with rlang and cli: formatted output, error chaining, and input validation.

targets-pipelines

Complex targets patterns: static branching, dynamic branching, hybrid patterns, and custom target factories.

tidy-evaluation

Programming patterns for data-masked functions in the tidyverse using {{}} and managing variable ambiguity.

tidymodels-overview

Overview of the tidymodels ecosystem for machine learning in R.

Commands

/r-lsp-diagnose

Diagnose R language server connectivity and configuration, and suggest fixes. Runs scripts/lsp-test-harness.sh, which exercises core LSP functionality and reports results as JSON (pass --fix to attempt repairs).