plain-language-writing

Two independent rule sets. A caller states which one applies — COMMS for text read once, live, in the moment; ARTIFACT for prose in a persistent file that gets re-read and reviewed. Don't cross-apply them; they're tuned for different reading conditions.

These rules operationalize standards with credible backing (ISO 24495-1, Kimble's plain-language work, EARS, INCOSE's Guide for Writing Requirements) but aren't each individually proven — treat them as the best currently-evidenced approach, not a guaranteed fix. See the research doc above for what's well-supported (the pronoun rule, COMMS #4 / ARTIFACT #3) versus reasonable-but-unproven. COMMS #5 (structure) has no empirical backing at all — an earlier version of it (A5) was dropped from the research for lack of evidence; it's kept here only as a stylistic default, not a proven rule.

COMMS — live agent-to-user text

Interview questions, confirmation/recap prompts, status updates, walkthrough questions, blocker/pause/stop messages.

Sentence length. Roughly 20 words or fewer. Split anything longer.
One clause per sentence. No "and/but/which" joining two distinct ideas.
No idioms, metaphors, or humor. Say it literally (not "let's dive in," "heads up," "ballpark").
No ambiguous pronoun. When a sentence introduces two things of similar weight, don't refer back with "it/this/that" — name the one you mean. Doesn't require eliminating every pronoun, only where more than one referent is plausible.
Causal/multi-step explanations use labeled structure, not a dense paragraph. When explaining a root cause or a multi-step process (why something failed, what happens as a result), break it into short ALL-CAPS section labels plus numbered steps or IF/THEN for branching logic. Doesn't apply to short interview questions or one-line status updates — those stay as plain sentences; forcing structure onto them reads as stilted.

Shape to follow: