Open-source Agent Skills for computational materials science and numerical simulation workflows.
Skills are folders of instructions, scripts, and resources that agents can discover and use. Write once, use everywhere.
Each skill includes:
SKILL.mdwith YAML frontmatter (name + description) and instructionsscripts/for deterministic workflowsreferences/for domain details loaded on demand
The agent discovers skills by name/description and loads the SKILL.md only
when the skill triggers. Scripts are executed directly for reliability.
| Skill | Description |
|---|---|
numerical-stability |
CFL analysis, von Neumann stability, stiffness detection |
numerical-integration |
Integrator selection, error norms, adaptive stepping |
linear-solvers |
Solver selection, preconditioner advice, convergence diagnostics |
time-stepping |
Time step planning, output scheduling, checkpointing |
differentiation-schemes |
Scheme selection, stencil generation, truncation error |
mesh-generation |
Grid sizing, mesh quality metrics |
| Skill | Description |
|---|---|
simulation-validator |
Pre-flight checks, runtime monitoring, post-flight validation |
parameter-optimization |
DOE sampling, optimizer selection, sensitivity analysis |
simulation-orchestrator |
Parameter sweeps, campaign management, result aggregation |
post-processing |
Field extraction, time series analysis, statistics, derived quantities |
- Examples for each skill in
examples/ - Comprehensive unit and integration tests in
tests/ - CI/CD pipeline for cross-platform testing (Python 3.10-3.12)
- Mention the skill by name in your request, or ask a task that matches its description.
- Run the scripts directly for reproducible outputs.
- Optionally install skills into your agent's global skills directory for reuse across projects.
Example:
Use numerical-stability to check a proposed dt for my phase-field run.
This repo follows the Agent Skills standard, originating with Anthropic,
and is designed to work with Claude Code, Codex-style agents, and any tool that
supports SKILL.md-based skills.
Run the full test suite:
python3 -m unittest discover -s testsBrowse and run examples:
cat examples/README.mdskills/ # Skill packages (SKILL.md, scripts, references)
examples/ # Runnable CLI examples
tests/ # Unit + integration tests
- Agent Skills standard: https://agentskills.io
- Reference implementation: https://github.com/agentskills/agentskills