ai-agent-skills

bioconductor AI Agent Skills

🏠 Home

🎯 Skills Index

⚙️ Installation

🤖 Agent Behavior

📋 Skill Standards

This is a repository of “Skills” intended to make AI coding agents (Claude Code, GitHub Copilot, Antigravity, Mistral, Aider, etc etc) perform more effectively and in alignment with community standards and best practices for Bioconductor users and developers. Its contents are written by humans and AI and vetted for accuracy and adherence to coding and analysis standards by human Bioinformaticians from the Bioconductor community.

What are AI Agent Skills?

Each “Skill” provides structured, Markdown-formatted instructions that teach AI agents about topics including

Package structures and conventions
Analysis workflows and best practices
Data types and their handling
Testing and documentation standards
Bioconductor-wide patterns

With these skills, AI agents can:

Create new skills and index them in SKILLS.md
Follow Bioconductor coding standards
Analyze R code and packages for security, efficiency, and code quality, and suggest improvements
Adhere to preferred bioinformatic workflows and statistical methods

Design Philosophy of these Skills

Workflow Orchestration vs. Code-Centric Snippets

These skills provide Workflow/Process Orchestration rather than simple code snippets. They define the intent, the multi-step workflow, and the domain knowledge required. This approach enables high autonomy and allows the AI to adapt to different project structures using its general reasoning capabilities.

The Role of Code Snippets as Guardrails: Code snippets serve as guardrails within a broader workflow. Use them specifically for highly specific APIs, standardized methodologies, or overcoming LLM anti-patterns. Use workflows for the “what” and “why”, and embed code snippets for the “how” only when exact syntax is required.

Centralized Registry vs. Decentralized Discovery

We use a Centralized Registry (SKILLS.md) so the LLM can understand how skills chain together through “Use Case Workflows” that orchestrate multiple skills.

Agent-Agnostic Design

Skills in this repository are agent-agnostic by design. They work across multiple AI agents through natural language invocation.

See AGENTS.md for detailed requirements and SKILL_STANDARD.md for technical format details.

Quick Start

Install the skills on your platform (see instructions/)
Ask your agent: “What bioconductor skills do I have?” to confirm
Use natural language to invoke skills (see SKILLS.md for examples)

Usage Examples

Create AI Instructions for an R Package:

"Create .github/instructions for this R package"

Analyze an R Package:

"Analyze this R package"

Create a New Skill:

"Help me create a new skill for [purpose]"

Repository Structure

ai-agent-skills/
├── README.md                      # This file
├── LICENSE                        # MIT License
├── AGENTS.md                      # Agent behavior standard (how discovery works)
├── SKILLS.md                      # Human-readable skill index (what skills exist)
├── SKILL_STANDARD.md              # Technical format spec (how to format skills)
├── CONTRIBUTING.md                # Contribution guidelines
├── CHANGELOG.md                   # Version history and release notes
│
├── skills/                        # All skills (flat structure)
│   ├── {skill-name}/              # Each skill in its own directory
│   │   ├── SKILL.md               # Skill documentation (required)
│   │   └── templates/             # Optional supporting files
│   └── ... (see SKILLS.md for complete list)
│
└── instructions/                  # Platform adapters (thin wrappers)
    ├── README.md                  # Adapter purpose and requirements
    ├── claude.md                  # Claude Code setup and shortcuts
    ├── copilot.md                 # GitHub Copilot setup and shortcuts
    └── antigravity.md             # Google Antigravity setup

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Support and Feedback

Issues: Report bugs or request features via GitHub Issues
Discussions: Ask questions in GitHub Discussions

License

MIT License - see LICENSE for details.

Acknowledgments

These skills are developed and maintained by the Bioconductor Agentic AI working group to make AI agents more efficient at developing and applying Bioconductor code.

This site is open source. Improve this page.