Ramda Skill

Use this skill when editing transformation-heavy TypeScript or JavaScript in this repo.

Principles

• Prefer Ramda when it makes data transformations more declarative and easier to scan.
• Avoid point-free code that hides types, branching, errors, or important domain names.
• Prefer R.pipe for readable left-to-right transformations when it avoids noisy intermediate variables.
• Plain loops are acceptable for complex branching, early exits, async control flow, or hot paths where Ramda obscures intent.
• Prefer a single reduce over filter plus map when one pass remains clear.

Preferred Operators

• Selection and shape: pick, pickBy, omit, props, path, pathOr, assoc, assocPath, evolve.
• Combining and merging: mergeLeft, mergeRight, mergeDeepRight, union, intersection, difference, symmetricDifference.
• Collections: groupBy, countBy, uniq, range, indexBy, partition, sortBy, ascend, descend.
• Predicates: any, all, anyPass, allPass, propEq, pathEq, includes.
• String/list transforms: split, splitAt, splitEvery, join.
• Fan-out and composition: juxt, applySpec, converge, pipe.
• Object utilities: invert, invertObj, toPairs, fromPairs, mapObjIndexed.

Type Discipline

• Keep undefined and null separate. Do not use R.isNil when absent and explicitly-null states need different behavior.
• Prefer typed helper functions at transformation boundaries so component and reducer code does not need casts.
• Reuse domain/API types when the transformation preserves the same shape.

Review Checklist

• Is the transformation clearer than the equivalent plain TypeScript?
• Are all intermediate domain concepts still named when naming helps understanding?
• Does the code preserve undefined versus null semantics when that distinction matters?