in-your-own-words

Minimize project documentation without losing truth.

Encodes domain rules in one structure, such as a state machine, typed model, or table, instead of scattered conditionals.

Finds what a change could break beyond its diff, and proves the one fact that makes it safe by running real code.

Concentrates validation and error handling at system boundaries, trusts internal types, and keeps business logic pure.

Cuts the layers and hidden state a reader must trace, collapsing one-caller wrappers and shrinking mutable scope.

Builds the script, codemod, or generator that does or proves the work, so a reviewer can rerun it.

Traces each symptom to the first place the system went wrong and fixes it there.

Settles core types, data structures, and shared state before writing logic.

Keeps the main context lean by routing bulk reading to subagents and keeping only their summaries.

Prefers deletion and the smallest change that solves the problem over new abstractions or layers.

Designs commands, lifecycle steps, and loops to reach the same end state after crashes, restarts, and retries.

Moves every caller to a new internal API and deletes the old one in the same wave.