matsci-agent-entry

Files (1)

• SKILL.md

SKILL.md (first lines)

# Matsci Agent Entry

## When to Use

Use this skill for materials-science tasks involving DFT, Quantum ESPRESSO,
Slurm/HPC execution, pseudopotentials, lattice constants, surface adsorption,
formation energies, convergence tests, EOS fitting, or BEEF uncertainty.

## Procedure

1. Identify which specialized matsci-agent skill is needed:
   - `matsci-agent:dft-qe-workflow` for QE inputs, convergence, EOS, lattice constants.
   - `matsci-agent:hpc-slurm-qe` for scheduler/resource/job handling.
   - `matsci-agent:surface-adsorption` for slab/adsorbate/adsorption energy tasks.
   - `matsci-agent:beef-uncertainty` for BEEF-vdW ensemble uncertainty.
   - `matsci-agent:convergence-debugging` for failed or inaccurate QE jobs.
   - `matsci-agent:sol27lc-protocol` for Sol27LC-style lattice-constant runs.
2. Load the relevant namespaced skill with `skill_view`.
3. Use the `matsci` toolset for parsing, state, convergence selection, dry-run
   Slurm scripts, and harness checks.
4. Default to dry-run unless the user explicitly authorizes live HPC execution.

## Pitfalls

- Do not mix exchange-correlation functionals when computing adsorption or
  formation energies.
- Do not store model/API keys in matsci-agent config.
- Do not submit real Slurm jobs unless live execution is explicitly authorized.

## Verification

Use `matsci_run_harness` for the closest materials-science check before real runs.

From chouyoudou/matsci-mcp, MIT.

More from chouyoudou/matsci-mcp