docs(prompts): tell Prime Claude when the KB beats a dictionary

[ddehilster]

A session reached for a .dict twice in one day to supply a fact that only one rule needed, and had to be shown otherwise both times. The primer never drew the distinction, so this adds it.

A dictionary changes what every rule sees; a knowledge base answers when asked. Loading a .dict into kb/user tags matching tokens in every pass for the whole run — a blast radius, and the right one when the analyzer is built on that vocabulary. An attribute like s= goes further still and changes tokenization: dicttok builds a typed node for those words, so plain _xALPHA rules stop matching them. For a fact needed in one rule, the KB is the instrument.

The pattern that goes with it: match on shape, validate in code — the rule is permissive, the loop is strict. A pass picking US state codes out of a list matches a run of two-letter uppercase tokens separated by commas, then loops calling a KB lookup (StateName()) on each candidate: two characters, uppercase, and known to the knowledge base. "multiple", "locations" and "in" each fail at least one of those three tests and drop out on their own, without the rule having to name them.

Placed in the "build on the libraries" rules of thumb, right after the bullet on wiring a .dict/.kbb pair into kb/user/ — that bullet describes the dictionary mechanism, so the qualification belongs next to it. The desc line picks up the new convention.

Prompt text only — no engine, analyzer, or library files touched.

🤖 Generated with Claude Code