Node API

npm install shibbo

import { check, rulesMarkdown, interfaceMarkdown, tells, resolveModel } from "shibbo";

const report = check(pageSource, "landing.html", { model: "gemini-4-argon" });
console.log(report.score, report.band);
for (const f of report.findings.filter((f) => f.counted)) {
  console.log(`${f.line}:${f.col} ${f.id} "${f.match}" → ${f.fix ?? f.name}`);
}

const block = rulesMarkdown({ model: "gpt-6.1-sol", limit: 10 });

check(text, file?, options?)

file sets the file kind from its extension ("-" means plain text), which decides the tells that run: prose, interface or code. Options:

Option	Type	Does
model	string | null	The writing model
kind	"markdown" | "text" | "html" | "css" | "markup-code" | "code"	Override the kind
disable	string[]	Tell ids to skip
only	string[]	Run only these tell ids
lexicon	boolean	false turns off measured phrases
domains	string[]	Override which domains apply

Returns a report.

Other exports

• rulesMarkdown({ model, domains, limit, phrases, agentLimit, clarity }) returns the rule block as Markdown. It opens with the Say it clearly section unless clarity is false, and lists agentLimit agent habits (default 6).
• interfaceMarkdown({ limit }) returns the rules for interface work: what to settle before building, then the commonest interface tells.
• rank(options) returns the ranked tells behind that block.
• tells(), models(), sources(), graphite() return the registry and measurements.
• resolveModel(id) maps any model id to a measured model, or null.