# What YES Means

> A plain-language decoder for one ballot measure at a time. Paste the ballot question and the measure's text; the
> browser numbers its sentences (S1, S2, ...; the question is Q), scores the reading level of the question and of the
> text (Flesch-Kincaid grade against an 8th-grade bar), flags negation traps such as the repeal of a ban (a double
> negative that makes YES mean "allow"), pulls out every dollar figure, percentage, rate, year and duration, spots
> constitutional amendments, bonds, taxes, repeals, new boards and expiry dates, glosses legal jargon, and draws a
> draft YES/NO fork from the wording - free, before you sign in. Two metered lanes run over the same measure: DECODE
> (wording plain / dense / double-negative; today's rules, what a YES vote changes, what a NO vote keeps, the money in
> the text, words to know, what the text does not say, where to check - every claim carrying an exact quote and the
> sentence it came from) and WEIGH (scope narrow / moderate / sweeping; who gains or loses decision power, rules that
> change, balanced trade-offs each side stresses, general patterns, questions to look up). Every quote in every reply
> is checked word for word against the pasted text, and every figure against the figures in the measure.

It explains only the text the voter pastes and never recommends a vote. Not affiliated with any election office,
campaign or party. The three built-in examples are fictional sample measures written for the app.

Derived from the SkillSafe agent skills @mengto/audit-verify-explain-grade-5 (audit, verify, explain simply - the
decode lane) and @rysweet/political-scientist-analyst (institutions, actors, power - the weigh lane).

## Pages

- [App](https://what-yes-means.skillsafe.ai/): paste a measure, read the wording check and the fork, run either lane.
- [API](https://what-yes-means.skillsafe.ai/api.html): the decode and weigh tasks over HTTP - input fields, reply
  contracts, worked examples, eight languages.
- [measurekit.js](https://what-yes-means.skillsafe.ai/measurekit.js): the free lane (sentence splitting, readability,
  traps, figures, quote verification); runs in node too.

## Contract in brief

- Input: `task` (decode | weigh), `measure_title`, `jurisdiction`, `election_date`, `ballot_question`, `sentences`
  (JSON-encoded `[{id, text}]`), `scan` (JSON-encoded page facts), `notes`, and for weigh an optional `decoded`.
- Output: one JSON object per task. Claims are `{text, cite[], quote}`; a quote must appear verbatim in a cited
  sentence.
- Reading grades are formula estimates. Figures shown are the measure's own; the app computes no costs.