# Kind-Lie Meter

> A light question check for adults with a small idea to try, at https://kind-lie-meter.skillsafe.ai/. Before asking friends, neighbours or colleagues about a stall, a class, a swap or a club, list up to eight questions you plan to ask. The page reads each one for how easily it invites a kind lie, a polite yes that tells you nothing, and draws every question as a polygraph trace. One metered run explains why each question draws the answer it does, rewrites the risky ones to ask about a real moment, says what a real answer sounds like and suggests one small closing ask.

URL: https://kind-lie-meter.skillsafe.ai/
API tutorial: https://kind-lie-meter.skillsafe.ai/api.html

## What it does

- Inputs: the small idea, who you plan to ask, up to eight questions in the order you would ask them, and an optional small step you hope they will take.
- Reads, riskiest first: fishing for praise (asks what they think of the idea), a pitch in disguise (tells them about it or leads them to agree), someday maybe (asks them to guess what they would do), usually-speaking (habits in general), life today (how things go for them now), a real last time (something that already happened). The page reads each question from cue words, first match winning in the order praise, pitch, someday, usually, a real last time, today; a question with no cue waits for you to pick its read.
- Free, in the browser: the reads, the polygraph chart (one trace per question, wilder for riskier reads) as SVG or PNG, the lamp (no question about their life = kind lies all round; every question about their life = straight answers likely; otherwise some kind lies likely), the question to start with (the riskiest one not already about a real last time; the earlier one on a tie), Markdown, CSV and JSON downloads, and saved readings with earlier verdicts for the same idea.
- Metered (one run on gpt-terra): per question, why it draws the answer it does and a rewrite that asks about the last time something actually happened (none for a question that already does), plus a headline, what to listen for and one small closing ask. The page reads every rewrite again with the same cue words and flags any that still invite a kind answer. One button puts the rewrites in place so the set can be read again.

## Contract

POST https://api.skillsafe.ai/v1/app-api/run with string fields: task = "meter", idea, people, questions (one "q1: text" line per question), hoped_ask (or "nothing listed"), facts (one line per question with its read and the cue the page found). The reply is one JSON object: {task, headline, questions: [{q, read, why, rewrite}], verdict, start_with, listen_for, closing_ask}.

## Notes

Friendly suggestions, not professional advice: the reads come from wording alone. Built on the SkillSafe skill @wondelai/mom-test (https://skillsafe.ai/skill/@wondelai/mom-test, MIT): talk about their life, not your idea; ask about specifics in the past, not opinions or the future; end with a small real commitment.