La interfaz está localizada; las descripciones y documentos de las Skills se traducen automáticamente con IA a su idioma. El código, los comandos, rutas y términos técnicos se mantienen en el original, y prevalece el documento del autor.
Explorar Skills
Productividad · SKILL.md

entries-thermodynamics-and-batteries

entries-thermodynamics-and-batteries

Prepare pymatgen computed entries, compatibility-corrected thermodynamics, battery electrode objects, and Borg VASP assimilation.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentación de la Skill

GitHub

name: entries-thermodynamics-and-batteries description: "Prepare pymatgen computed entries, compatibility-corrected thermodynamics, battery electrode objects, and Borg VASP assimilation." disable-model-invocation: true metadata: disco-role: operating license: NOASSERTION

Entries, Thermodynamics, and Batteries

Use this sub-skill when a task mentions ComputedEntry, ComputedStructureEntry, compatibility corrections, MP 2020 corrections, GGA/GGA+U/r2SCAN mixing, phase-diagram-ready entries, entry grouping, insertion or conversion batteries, voltage pairs, BatteryAnalyzer, VaspToComputedEntryDrone, or BorgQueen.

Route

  • Read references/api-reference.md for imports, signatures, metadata expectations, and object relationships.
  • Read references/workflows.md for recipes to prepare entries, preserve correction metadata, group entries, construct battery objects, and assimilate VASP outputs.
  • Read references/troubleshooting.md for predictable failures involving missing metadata, filtered entries, oxidation states, incomplete VASP output, DFT mixing, and working-ion assumptions.
  • Run scripts/entry_battery_smoke.py for a safe in-memory import/readiness check; it does not read user calculation directories, fixtures, credentials, or network resources.

Boundaries

  • Use this sub-skill after entries are already local or in memory. Live Materials Project, COD, OPTIMADE, credentials, field names, pagination, and network errors belong to ../external-data-access/SKILL.md.
  • Use this sub-skill for entry and battery Python APIs. pmg analyze, persistent configuration, and other CLI syntax belong to ../cli-and-configuration/SKILL.md.
  • Use this sub-skill for phase-diagram-ready entry preparation, then hand core hull, reaction, Pourbaix, surface, and interfacial interpretation to the owning analysis APIs or ../surfaces-interfaces-and-electrochemistry/SKILL.md.
  • Treat Borg assimilation as user-directory-specific and path-sensitive. Do not scan broad filesystems unless the user provides the calculation root and accepts the scan.

Default Approach

  1. Identify the entry source: in-memory values, saved Monty JSON, local VASP outputs, or an external-data handoff.
  2. Preserve provenance before corrections: composition, total energy in eV per entry, parameters["run_type"], parameters["hubbards"], parameters["potcar_spec"] or parameters["potcar_symbols"], optional parameters["software"], structure, volume, and trusted oxidation-state/anion hints.
  3. Apply the compatibility or mixing scheme that matches the calculation provenance before building phase diagrams, reactions, or battery electrodes.
  4. Keep exploratory processing non-destructive with inplace=False, and debug filtering with on_error="warn" or on_error="raise".
  5. For battery workflows, verify whether the task is a redox-capacity estimate, an insertion path with a stable topotactic framework, or a conversion path requiring a complete phase diagram.
Skills guardadas