Agentic Skills — Documentation skill for Claude Code

Agentic Skills for R & Bioinformatics

Opinionated agent skills for single-cell RNA-seq analysis, R package development, and scientific documentation. Built for bioinformatics workflows using Seurat, Bioconductor, tidyverse, and Quarto.

These skills encode specific conventions, patterns, and quality standards that AI coding agents follow when working on your projects. They are deliberately opinionated — fork and adapt them to your own workflow.

Supported Agents

This repository works with both **Claude Code** and **Codex**. Each tool discovers skills differently:

• Claude Code reads SKILL.md files directly from ~/.claude/skills/, using YAML frontmatter (name, description) to determine when to load each skill.
• Codex reads SKILL.md files from ~/.codex/skills/ and uses agents/openai.yaml metadata for interface display names, descriptions, and default prompts.

Both tools share the same skill content (`SKILL.md`). The `agents/` subdirectories are Codex-specific and harmless to Claude Code — they are simply ignored.

What Are Skills?

Skills are markdown files (`SKILL.md`) that teach an AI agent domain-specific conventions. When the agent detects a relevant task (e.g., editing a `.qmd` file), it loads the matching skill and follows its rules. Skills live in the agent's skills directory as subdirectories, each containing a `SKILL.md` file and optionally an `agents/openai.yaml` for Codex metadata.

Codex Agent Metadata

Each skill may include an `agents/openai.yaml` file that provides Codex-specific UI metadata:

• display_name — how the skill appears in Codex's interface
• short_description — tooltip text shown when browsing skills
• default_prompt — the prompt template Codex uses when invoking the skill

Claude Code does not use these files; it reads the YAML frontmatter in `SKILL.md` directly. The `agents/` directories are committed to the shared repo so both tools stay in sync, but they only affect Codex behavior.

Skills