Deduce Mine: A Minefield That Deals Itself
Sweep a field where the mines are not planted when the game starts, in one small offline HTML file: every reveal asks a small constraint solver which placements still fit the clues and flags, then deals one. Provably-safe cells are guaranteed empty, provably-deadly ones guaranteed mines, ambiguous ones a fair coin — so a Hint toggle genuinely glows forced cells.
Bring the grid and the mine count, the solver rules and the hint, and optionally the flag mode, the undo, the palette and the UI text. The Skill returns a downloadable Skill (the deduction minesweeper project, a customize guide, headless verification) and one playable example field.
How it works
Your coding agent copies the working game and follows the customize guide: the deferred-deal loop, the constraint solver, provably-safe and provably-deadly reads, the Hint glow and the synth audio.
The published example deals mines lazily: nothing is planted at start, and every reveal asks the solver which placements still fit the clues and flags, so provable cells are guaranteed and ambiguous ones are a fair coin.
The game is played headlessly under the sandbox policy: menu, a safe first reveal, more reveals, the Hint glow, flag and unflag, a solver-certified safe cell, undo, pause and mute, the Chinese UI and a phone reveal.
Frequently asked questions
Does it call a model or need a key?
No. The Skill gives your agent a working game, a customize guide and verification; the game itself makes no network calls and calls no model.
Why is the demo sandboxed?
The frame has an opaque origin and no network access, so the scene can’t read anything from this page or reach the internet.
Is the game bilingual?
Yes. Both the page and the game’s own UI switch between English and Chinese.
What are the visuals and audio made of?
All of it is code: Canvas 2D draws the grid, clues and glow, and WebAudio synthesizes the ticks. No image, font or audio files ship with it.
Does it work on mobile?
Yes: tap to reveal, flag mode to mark. Verified under touch emulation, not on real devices.
Why is there no online generation?
Building a new scene needs a model key inside the sandbox, which the backend doesn’t support yet. Run the Skill with your own agent for now.