Code Construction
Overview
Four leaves cover the craft of producing new code: how a unit is written (naming, errors, gates), the order and size of the steps it lands in, the throwaway probes that answer a question before real work, and the in-repo documentation that ships with the change.
When to use
- Writing new functions, modules, error paths; setting up formatter, linter, type gates.
- Landing work in small green steps; building a walking skeleton; running a Parallel Change.
- Running a time-boxed spike or prototype to answer one question.
- Writing docstrings, comments, or a generated API reference in the repo.
When not to use
- Slicing the spec into tasks or estimating it — outside this skill.
- Refactoring or simplifying existing working code as the main goal — outside this skill.
- Choosing what to test across the whole suite, or branch/merge policy — outside this skill.
- User-facing prose such as READMEs, tutorials and release notes.
Precedence
The deepest loaded instruction governs its uniquely-claimed scope; every
scope:string is claimed by exactly one file in this tree; no leaf may widen a prohibition declared by an ancestor — leaves narrow, never override upward.
Subskill index
| Leaf | Task | Trigger |
|---|---|---|
subskills/construction.md |
Write new code | "implement this" |
subskills/incremental-delivery.md |
Small verified steps | "build it step by step" |
subskills/spikes-prototypes.md |
Time-boxed throwaway experiment | "spike this for two days" |
subskills/code-docs.md |
Comments, docstrings, generated reference | "document this function" |
See also
subskills/construction.md— the unit of code.subskills/incremental-delivery.md— the order the units land in.subskills/spikes-prototypes.md— the throwaway probe before real work.subskills/code-docs.md— the prose that ships beside the code.