nf-core-module-dev

A Claude Code plugin that gives your coding agent specialised skills for creating, testing, and documenting nf-core Nextflow modules.

How it works

When you ask Claude to build an nf-core module, it doesn't just start writing code. The nf-module-manager skill orchestrates a structured pipeline:

1. It delegates main.nf and environment.yml to nf-module-dev, which researches the tool, scaffolds the module, and populates it following current nf-core conventions.
2. It then launches nf-test-expert and nf-secretary in parallel — one writes and runs the nf-tests, the other writes the meta.yml.
3. Once the snapshot exists, the meta.yml is linted. Errors are routed back to the right agent automatically (up to 3 retries per agent).
4. You get a final report: module files, test results, lint status, and any warnings.

Each agent is a specialist:

Agent	Responsibility
nf-module-dev	main.nf, environment.yml
nf-test-expert	nf-tests, snapshots
nf-secretary	meta.yml, linting

Because the nf-module-manager runs as a skill in your main session (not a hidden sub-agent), you can watch every step as it happens.

Installation

claude plugin marketplace add vagkaratzas/nf-core-module-dev
claude plugin install nf-core-module-dev@vagkaratzas

Then start a new session. The plugin bootstraps automatically.

Usage

For a complete end-to-end module build, just describe what you want:

"Create an nf-core module for bwa-mem2 mem"

Claude will invoke the nf-module-manager skill and orchestrate the full pipeline.

For targeted work, agents can be invoked directly:

"Write an nf-test for the clipkit module"
"Fix the meta.yml lint errors in samtools/sort"
"Update eggnogmapper to the latest bioconda version"

What's inside

Skills

• nf-module-manager — end-to-end module orchestration with parallel agents and automatic error attribution
• using-nf-core-module-dev — session-start bootstrap that tells Claude when to reach for each agent

Agents

• nf-module-dev — researches tools, scaffolds modules, populates main.nf and environment.yml following current nf-core style. Stops and waits for you if a bioconda/container environment can't be auto-detected.
• nf-test-expert — writes nf-tests calibrated to current community style, generates and verifies snapshots, supports singularity/docker/conda profiles
• nf-secretary — writes and lints meta.yml, handles topic-based versions, EDAM ontologies, and all known schema edge cases

Requirements

• Claude Code
• A local clone of nf-core/modules
• nf-core CLI installed and on your PATH
• nf-test installed and on your PATH
• Singularity, Docker, or conda for running tests

Contributing

Agents carry their domain knowledge as embedded reference sections — no separate memory files. To contribute a fix or new pattern:

1. Fork the repository
2. Edit the relevant ## Reference: ... section in the agent .md file
3. Submit a PR

Releasing

Version numbers are kept in sync across plugin.json and marketplace.json using the included bump script:

# Check versions are in sync
scripts/bump-version.sh --check

# Bump to a new version
scripts/bump-version.sh 1.1.0

## License

MIT