# Ticks and Crosses

> A logic-grid puzzle maker for your own small facts. Type three or four short lists - who, which mug, which
> drink - with three to five items each, in matching order: the first items belong together, the second items
> together, and so on. The page deals true clues of four kinds (goes with, not, either-or, one-and-the-other),
> keeps only the ones its deduction-only solver needs, and hands you a puzzle with exactly one answer that never
> needs a guess. Solve it on a classic staircase grid of ticks and crosses. A metered run words the clues in a
> voice you pick. Built on the SkillSafe skill @guia-matthieu/thought-based-reasoning.

https://ticks-and-crosses.skillsafe.ai/

## What is free
- Dealing the puzzle from your lists, New clues (a different set for the same answer), and the playing grid: tap a cell for a cross, again for a tick; auto-cross fills the rest of a row and column after a tick.
- Check my ticks (right, wrong and crossed-by-mistake counts), Show the answer, a readout of what your ticks say so far, and up to six step hints that follow the solver's first moves.
- Items that look like an email, phone number, handle or link are kept out of the puzzle and never sent.
- Three sample puzzles replay their saved wording at no cost.
- Downloads: the puzzle sheet (SVG, blank grid plus clues), print, the puzzle to hand over (copy, or share on a phone: the lists and clues, no answer), and the puzzle with its answer (Markdown, JSON).

## The metered task
One task, `dress`, on gpt-terra, metered per run. The input is string fields: `task`, `voice` (`plain`, `cosy`, `grand` or `brisk`), `setting`, `lists`, `clues` (`C1 (kind): plain clue.`), `clue_count`, `steps` (`H1 (look at clue N): A goes with B.`) and `step_count`. The reply is one JSON object: `title`, `intro`, `clues` (`{id, text}` per clue) and `hints` (`{id, text}` per step).

The page checks every reply: each worded clue must name its own items and no others and keep its kind (a negative only on not-clues, the or in either-or, one and the other in the pairs), with no added number words or internal ids; each hint must name its placement and point at its own clue; the title and introduction must name no item. Any clue that fails keeps the page's plain wording, so the puzzle on screen is always the one that was dealt. The page makes puzzles; it does not check facts about the people or things you type.

## Links
- API tutorial: https://ticks-and-crosses.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@guia-matthieu/thought-based-reasoning (guia-matthieu/clawfu-skills)