Game review

Dragonsweeper

8.5 / 10

Developer
Daniel Benmergui
Publisher
Daniel Benmergui (itch.io: danielben)
Released
1 January 2025
Platforms
Web browser (desktop and mobile)
Genre
Puzzle / Roguelike
Monetisation
Free, name-your-price on itch.io
Players
Single-player only

The short version

Dragonsweeper's premise is a single clean idea: every Minesweeper number is now a monster's hit points, and stepping on one doesn't end the game, it starts a fight your tiny knight might lose. That's enough of a twist to turn a 35-year-old logic puzzle into something that plays like a roguelike, and it earns its viral 2025 reputation and 4.9-star, 2,000-plus-rating average on merit — tested directly, the deduction is still real Minesweeper deduction, and the combat layer adds a second axis of risk that the genre never had. For this shelf, the good news is straightforward: reveal-a-tile-with-a-tap, the entire core loop, was confirmed working identically on a simulated touch phone and a desktop mouse, tap for tap. A more advanced hold-to-mark gesture exists in the code for bookkeeping already-seen threats, built on the same input-agnostic logic as everything else, though this review could not independently trigger it through automated touch testing — worth flagging honestly rather than claiming more certainty than the testing supports.

Biggest strengths

  • The core interaction — tap a tile to reveal it, tap a revealed monster to fight it — was confirmed working via real touch events on a simulated phone, not inferred from the presence of a `touchstart` listener; the same action produced identical results with a mouse on desktop
  • The central idea is a genuinely fresh twist on a design most people assumed had nothing left to say: numbers become monster health, "flags" become tactical retreats, and the deduction puzzle now has a survival-horror undercurrent, since guessing wrong can mean death rather than just a restart
  • `gamelib.js` wires touch input through the identical code path as mouse input at the engine level — `touchstart`/`touchmove` update the same position variables `mousemove` does, and `touchend` calls the same interaction function `onclick` does — rather than treating touch as an afterthought

Biggest weaknesses

  • The game ships with no licence anywhere — no LICENSE file, no statement on the itch.io page beyond a name-your-price download of the source for buyers to inspect. The developer also sells other commercial games (Storyteller, on Steam), and this shelf is aware of at least one community report of an unauthorised mirror the developer did not sanction — reason for extra care, not a reason to skip the shelf's existing unlicensed-work policy, but worth being transparent about: this copy carries the developer's name, a link to the original, and a same-day takedown offer, precisely because no licence was ever granted
  • A secondary hold-to-mark gesture exists in the code for tagging already-revealed threats, built on the same shared touch/mouse state as everything else, but repeated attempts to trigger it through automated long-press testing on this shelf's phone profile did not conclusively confirm it in a screenshot — most likely a limitation of synthetic touch timing rather than a real defect, but not verified either way
  • It is genuinely difficult, in the way roguelikes are meant to be: several tiles revealed during testing ended a run immediately, and new players should expect short first attempts

Who it's for: Anyone who likes Minesweeper, anyone who likes small roguelikes, and especially anyone who likes both and has never seen them combined this cleanly

Our recommendation: One of the strongest candidates reviewed for this shelf — a genuine, viral, well-earned hit

8.5 / 10

The elevator pitch for Dragonsweeper sounds almost too simple to be worth building: it's Minesweeper, except the numbers are monsters instead of proximity counts, and clicking one starts a fight instead of ending the game. Played directly, that one change turns out to be enough to make the entire genre feel new again. You're still doing the same deduction work — reading numbers, inferring what's safe, planning which tile to risk next — but every guess now carries a second kind of stakes. A wrong tile isn't just "game over, try again," it's "your knight just took a hit from something with more health than he does," and the tension of that lands every time.

The core loop works on a phone, confirmed rather than assumed

The reason this belongs on a shelf built around phone playability isn't a guess from reading the source — it's confirmed by driving the actual page through real touch events. gamelib.js sets up touchstart and touchmove listeners that write to the exact same mouseScreenX and mouseScreenY variables the desktop mousemove handler uses, and touchend calls activatePlayerInteraction(), the identical function a desktop click triggers. That's not touch bolted onto a mouse game; it's one interaction model with two input sources feeding into it equally. Testing this on a simulated 390×844 touch phone, a tap on an unrevealed tile correctly revealed it, and a tap on an already-revealed monster correctly triggered combat — in one run, walking directly into a dragon's tile this way ended in "Torched by the Dragon," which is exactly the outcome the same action produces with a mouse on desktop. The core game is completely playable with a thumb.

There's a second, more advanced gesture in the code worth naming honestly rather than glossing over: holding a tile for around a third of a second is meant to cycle through status markers for tagging monsters you've already seen, useful bookkeeping on a board that gets crowded. Reading game.js, this is built on the same mousePressed state both touch and mouse populate identically, so there's every reason to expect it works the same way on a real finger as it does in the code. Repeated attempts to independently confirm it through this shelf's automated long-press testing didn't produce a clear screenshot of the marker menu opening, most likely a quirk of how synthetic touch timing interacts with the game's per-frame hold-duration check rather than a real gap — but this review would rather say "not independently confirmed" than claim certainty it doesn't have. It's also a convenience feature, not a requirement: the entire game is completable using only tap-to-reveal and tap-to-fight.

Why it went viral

None of the above would matter if the underlying game weren't good, and it clearly is. The production values are a big part of it — a small, expressive pixel-art knight, monster portraits with real character, a HUD that communicates health and progress at a glance — but the design is what earns the 4.9-star average across more than two thousand ratings. Turning "numbers near a mine" into "hit points on a monster" is the kind of idea that sounds obvious in retrospect and apparently wasn't obvious to anyone for the genre's first three and a half decades. It rewards the exact same careful, patient reading that made the original popular, while adding a layer of risk-management that gives every session its own shape.

The licensing situation, stated plainly

There is no licence attached to this game anywhere — not in a file, not on the itch.io page beyond a paid-or-free download of the source for people who buy it to look at. The developer also sells other commercial work, and this shelf is aware of at least one community mention of an unauthorised mirror the developer didn't sanction, which is worth taking seriously as a signal rather than ignoring. That context doesn't change the underlying facts this shelf's current policy is built around — no licence was granted, but nothing explicitly restrictive was stated either — so this is hosted the same way other unlicensed-but-not-restricted work has been this cycle: credited to the developer by name, linked to the original itch.io page, and offered a same-day takedown on request. That's a meaningfully different posture than an anonymous, uncredited mirror, and it's the right way to host something this good without a licence in hand.

Where that leaves it

An 8.5 reflects a genuinely inventive, extremely well-received small game whose core interaction was tested and confirmed to work cleanly with a thumb, held back only slightly by one unconfirmed secondary gesture and the honest complexity of its licensing situation. This is exactly the kind of thing this shelf exists to surface.

Pros and cons

Pros

  • The core interaction — tap a tile to reveal it, tap a revealed monster to fight it — was confirmed working via real touch events on a simulated phone, not inferred from the presence of a `touchstart` listener; the same action produced identical results with a mouse on desktop
  • The central idea is a genuinely fresh twist on a design most people assumed had nothing left to say: numbers become monster health, "flags" become tactical retreats, and the deduction puzzle now has a survival-horror undercurrent, since guessing wrong can mean death rather than just a restart
  • `gamelib.js` wires touch input through the identical code path as mouse input at the engine level — `touchstart`/`touchmove` update the same position variables `mousemove` does, and `touchend` calls the same interaction function `onclick` does — rather than treating touch as an afterthought
  • Clean under the hood: no third-party analytics anywhere in the source, and the only outbound link in the code is a plain, visible click-through to the developer's other commercial game, not a hidden tracking call
  • A 4.9-star average across more than 2,100 itch.io ratings, and a design distinctive enough that "went viral" is a fair description rather than marketing language

Cons

  • The game ships with no licence anywhere — no LICENSE file, no statement on the itch.io page beyond a name-your-price download of the source for buyers to inspect. The developer also sells other commercial games (Storyteller, on Steam), and this shelf is aware of at least one community report of an unauthorised mirror the developer did not sanction — reason for extra care, not a reason to skip the shelf's existing unlicensed-work policy, but worth being transparent about: this copy carries the developer's name, a link to the original, and a same-day takedown offer, precisely because no licence was ever granted
  • A secondary hold-to-mark gesture exists in the code for tagging already-revealed threats, built on the same shared touch/mouse state as everything else, but repeated attempts to trigger it through automated long-press testing on this shelf's phone profile did not conclusively confirm it in a screenshot — most likely a limitation of synthetic touch timing rather than a real defect, but not verified either way
  • It is genuinely difficult, in the way roguelikes are meant to be: several tiles revealed during testing ended a run immediately, and new players should expect short first attempts
  • No difficulty settings; the challenge is fixed by design

Category scores

Category scores are the reviewer's read on each area. They inform the overall score but do not average into it — the number at the top of the page is a judgement, not a calculation.

  • Gameplay 9 / 10
  • Visuals 8.5 / 10
  • Originality 9 / 10
  • Replayability 8 / 10