CatalystNeuro Claude Code Skills

Public repository of Claude Code skills for neurophysiology research.

Installation

Prerequisites

1. Install Claude Code
2. Ensure you have an active Anthropic API key or Claude Pro/Max subscription

Adding Skills

Register this repository as a Claude Code plugin marketplace:

/plugin marketplace add catalystneuro/claude-skills

Then install individual skills:

/plugin install analyzing-dandi-datasets@catalystneuro-skills
/plugin install using-nemos@catalystneuro-skills
/plugin install using-pynapple@catalystneuro-skills
/plugin install nwb-convert@catalystneuro-skills

Manual Installation (Alternative)

If you prefer to install skills without the marketplace, you can clone this repo and add a skill directly:

git clone https://github.com/catalystneuro/claude-skills.git ~/claude-skills

Then in Claude Code:

/skill add ~/claude-skills/nwb-convert

Verifying Installation

After installing a skill, you can verify it's available:

/skills

This will list all installed skills. You should see the skill name in the list.

Available Skills

nwb-convert

Convert neurophysiology data to NWB format and publish on DANDI. This skill acts as an expert NWB conversion specialist, guiding you through the entire conversion process:

1. Experiment Discovery - Understand your data modalities, recording systems, and file organization
2. Data Inspection - Automatically inspect files to identify formats, channels, and structure
3. Metadata Collection - Gather required NWB metadata (subject, session, devices, electrodes)
4. Synchronization - Analyze and plan temporal alignment across data streams
5. Code Generation - Generate a complete, pip-installable conversion repo using NeuroConv
6. Testing & Validation - Run conversions, validate with NWB Inspector, fix issues iteratively
7. DANDI Upload - Organize and upload validated NWB files to the DANDI Archive

Supported modalities:

• Extracellular electrophysiology (SpikeGLX, OpenEphys, Intan, Blackrock, Neuralynx, Plexon, TDT, Axona)
• Spike sorting (Kilosort, Phy, SpykingCircus, MountainSort, YASS, Combinato)
• Calcium imaging (ScanImage, Scanbox, Bruker, MicroManager, Miniscope, Hamamatsu)
• Segmentation (Suite2p, CaImAn, EXTRACT, CellPose)
• Behavior (DeepLabCut, SLEAP, FicTrac, video, custom formats)
• Intracellular electrophysiology (ABF, WinWCP)

Usage:

/nwb-convert /path/to/your/data

Or simply describe what you want to convert:

• "I have SpikeGLX recordings with Kilosort sorting and behavioral data from a VR task"
• "Convert my two-photon calcium imaging data with Suite2p segmentation to NWB"
• "Help me publish my electrophysiology dataset on DANDI"

Knowledge base includes:

• 68 NeuroConv interface specifications
• Canonical conversion repo structure (cookiecutter template)
• Patterns from ~20 real CatalystNeuro conversion repos
• NWB best practices distilled from NWB Inspector

analyzing-dandi-datasets

Analyze neurophysiology datasets from the DANDI Archive. Load NWB files with streaming access, use Pynapple for data inspection, and create analysis pipelines for neural phenomena like directional tuning, place cells, and population dynamics.

Requires: neurosift-tools MCP

using-nemos

Fit Generalized Linear Models (GLMs) to neuroscience data using the NeMoS Python package. Covers:

• Basis functions (BSpline, RaisedCosineLog, CyclicBSpline, Eval vs Conv)
• Observation models (Poisson, Gaussian, Gamma, Bernoulli)
• Regularization (Ridge, Lasso, GroupLasso)
• Single-neuron and population GLMs
• Functional connectivity and coupling filter analysis
• Cross-validation and model selection with scikit-learn
• Calcium imaging with Gaussian GLMs

using-pynapple

Analyze neurophysiology time series using the pynapple Python package. Covers:

• Core data structures (Ts, Tsd, TsdFrame, TsdTensor, TsGroup, IntervalSet)
• Time series manipulation (restrict, count, smooth, interpolate, bin_average, derivative)
• Metadata management and neuron filtering
• Tuning curves (1D, 2D, n-dimensional)
• Bayesian and template decoding
• Signal processing (filtering, wavelets, Hilbert phase extraction)
• Correlograms and perievent analysis

Usage

After installing a skill, Claude Code will automatically use it when relevant. You can also invoke skills directly with their slash command:

• /nwb-convert - Start an NWB conversion workflow
• "Load this NWB file and compute head direction tuning curves" - triggers using-pynapple
• "Fit a Poisson GLM with spike history basis" - triggers using-nemos
• "Find a DANDI dataset with hippocampal place cells" - triggers analyzing-dandi-datasets

The using-nemos and using-pynapple skills cross-reference each other since NeMoS workflows typically use pynapple for data preparation.

Skill Architecture

Each skill directory contains:

• SKILL.md - Main skill definition with frontmatter (name, description, tools) and instructions
• Reference files - Knowledge bases, patterns, and examples that the skill consults

The nwb-convert skill is the most complex, with 7 phase-specific instruction files and 4 knowledge base files covering interfaces, repo structure, conversion patterns, and NWB best practices.

License

MIT