Agent Skills

The ALCHEMI Toolkit ships a set of agent skills — concise instruction files that AI coding assistants (Claude, Copilot, Cursor, etc.) can load to get up to speed with the nvalchemi API without lengthy context-gathering.

Skills live in the repository under .claude/skills/.

Installing skills

Inside a repository clone — nothing to install. Claude Code discovers .claude/skills/ automatically; other agents are routed to the right SKILL.md by the table in the repository’s AGENTS.md.

Outside a clone — copy the skill folders from .claude/skills/ into your project’s skills directory, or your user-level one (e.g. ~/.claude/skills/) if you work with nvalchemi across many checkouts.

Available skills

Skill	Description	Related user guide
nvalchemi-data-structures	How to use AtomicData and Batch for representing atomic systems and batching them for GPU computation.	AtomicData and Batch
nvalchemi-data-storage	How to write, read, compose, and load atomic data using the composable Zarr-backed storage pipeline (Writer, Reader, Dataset, MultiDataset, DataLoader).	Data Loading Pipeline
nvalchemi-zarr-perf	How to tune Zarr-backed Reader, Dataset, MultiDataset, and DataLoader throughput with fused reads, validation skipping, pinned memory, and benchmark sweeps.	Read performance tuning
nvalchemi-model-wrapping	How to wrap an arbitrary MLIP using the BaseModelMixin interface to standardize inputs, outputs, and embeddings.	Models: Wrapping ML Interatomic Potentials
nvalchemi-training-api	How to configure, run, extend, and debug TrainingStrategy workflows.	Training
nvalchemi-loss-api	How to use built-in loss terms, compose training objectives, and implement custom loss functions.	Losses
nvalchemi-fine-tuning	How to configure fine-tuning workflows and adapt pretrained checkpoints through the CLI or API.	Fine-Tuning Pretrained Models
nvalchemi-dynamics-api	How to configure and run dynamics simulations, compose multi-stage pipelines (FusedStage, DistributedPipeline), use inflight batching, and manage data sinks.	Dynamics: Optimization and Molecular Dynamics
nvalchemi-dynamics-implementation	How to implement a dynamics integrator by subclassing BaseDynamics and overriding pre_update() and post_update().	Dynamics: Optimization and Molecular Dynamics
nvalchemi-dynamics-hooks	How to use and write dynamics hooks — callbacks that observe or modify batch state at specific points during each simulation step.	Hooks
nvalchemi-reporting	How to add observability with ReportingOrchestrator, RichReporter, TensorBoardReporter, and the dynamics LoggingHook.	Reporting