| name | abacus-librpa-abfs-orbital |
| description | Obtain, select, validate, or document ABFS_ORBITAL inputs for ABACUS + LibRPA workflows. Use when the user asks how to get .abfs files, how to write ABFS_ORBITAL entries in STRU, how to build or source auxiliary bases, or when the user is teaching a new ABFS_ORBITAL recipe that should be saved for reuse. |
ABACUS + LibRPA ABFS_ORBITAL
Use this skill when the task is specifically about auxiliary-basis files and ABFS_ORBITAL, not the full GW route.
Core workflow
- Identify which of these intents applies:
- the user already has
.abfs files and wants to wire them into STRU
- the user wants to know how to obtain
.abfs files for a given system
- the user wants to validate whether existing
.abfs files match the NAO / pseudopotential setup
- the user is teaching a new
ABFS_ORBITAL method that should be stored as a reusable playbook
- Read
references/abfs-orbital-recipe.md before giving procedural guidance.
- Keep the source of truth explicit:
- where the
.abfs files come from
- which ABACUS / LibRPA lane they are meant for
- whether the method is generic, element-library based, or host-specific
- Treat provided
.abfs files as authoritative assets unless the user asks to regenerate them.
- When the user is teaching a new method, update
references/abfs-orbital-recipe.md with the exact steps, not a paraphrased summary.
What to record when the user teaches a method
Capture these items exactly:
- prerequisites: code branch, binaries, scripts, host assumptions
- required inputs:
.orb, .upf, STRU, element list, cutoff/radius assumptions
- exact commands
- expected output filenames and naming rules
- how the resulting filenames map into
ABFS_ORBITAL lines in STRU
- validation checks
- known failure modes or incompatible combinations
If any of those are still unknown, leave a clear TODO: line instead of inventing details.
Validation rules
Whenever checking an ABFS_ORBITAL setup, verify at least:
- each species in
STRU has the intended .abfs file
- file naming and species mapping are unambiguous
- the method is consistent with the user's shrink / PCA lane
- if the workflow mixes different NAO cutoff radii across species, record that explicitly because it can matter for debugging
Output style
When answering from this skill, report:
- source of the
.abfs recipe
- exact
ABFS_ORBITAL entries or exact next command
- what was verified
- any unresolved assumptions