Stolen Sword

Stolen Sword's pitch is one sentence long and it is not lying to you: it is a one-finger action platformer where a single press-drag-release gesture handles aiming, jumping, and attacking at once. Ian Chiao built it for js13k 2020 — the competition that caps entries at 13 kilobytes zipped — and the whole game, four stages included, ships as one 31KB HTML file that unpacks from that budget. What makes it worth a look six years later isn't the file-size trick, though; it's that the control scheme it had to invent to fit a whole combat system into one gesture turns out to be genuinely good, once you understand what it is actually asking you to do.

The gesture is a slingshot, not a flick — and that matters

Reading src/modules/interaction.js settles the registry's claim outright: the code binds mousedown/mousemove/mouseup and touchstart/touchmove/touchend to the exact same three handler functions, and the old keydown/keyup listeners sit commented out above them. This was built touch-first and never had its symmetry bolted on afterward — there is no separate "mobile" code path to fall out of sync with the real one.

But the gesture itself needs stating plainly, because nothing in the game states it: this is a pull-back-and-release slingshot, not a flick in the direction you want to travel. The release velocity in state.js is calculated as the vector from your release point back to your press point — press near the top of the screen, drag down and release, and the character launches up, the same way pulling a slingshot's pouch backward sends the shot forward. It is precisely the Angry Birds convention, and once you know that, aiming a slash at an enemy above and to the right is just "press up-right, drag down-left, let go." Verified directly: playtesting on a simulated 390×844 touchscreen with that gesture launched the character straight into the first enemy, cleared it, and advanced the fight into the game's second wave, with zero console errors and zero page errors across every test run. The mechanic works, and dropping the whole screen to 5% time speed for the duration of the press — easing back to normal the instant you let go — turns what would otherwise be a twitchy mid-air aim into something genuinely readable. It's a cheap effect and a smart one.

The problem is that the game never tells you any of this in words. The only teaching is a faint, once-only dashed line animated across the first two waves, showing the correct pull-and-release motion — easy to miss, and easy to misread as decoration rather than instruction. A first-time player's instinct is to flick in the direction they want to go, the same way Fruit Ninja or a dozen other touch games work, and that instinct is exactly backward here. Get the gesture wrong and the character does nothing useful; a player who doesn't stumble onto the reversal within a try or two could reasonably conclude the controls simply don't work, which is a real risk for a touch-only shelf whose whole premise is that the controls have to work.

Four short stages that don't repeat themselves

For a 13-kilobyte competition entry, the source shows real variety rather than one loop copy- pasted with bigger numbers. Stage one is the closest thing to a "normal" combat gauntlet: eight small waves that escalate from a single stationary enemy, to a pair, to a six-enemy swarm with staggered spawn timing, to a twelve-node chained enemy that follows its own head around the screen. Stage two abandons combat waves entirely for one long vertical climb through six moving, untouchable hazards — pure platforming precision, no sword involved. Stage three changes the rules again: lower gravity, an underwater section with its own swimming animation, and "shell" enemies that actively defend and only take real damage once you've landed a second hit past their guard. Stage four closes with one sprawling encounter of roughly twenty enemies scattered across a much taller vertical space, including a ten-link chained enemy that has to be threaded rather than simply charged.

That structural variety is a genuine strength — this is not "one mechanic, four palette swaps." But the whole thing is also over fast: four stages, around fifteen hand-placed encounters total, clearable in ten to fifteen minutes by a player who has the gesture down. It ends on a literal "Still can't find him... (To be continued)" caption, and this build never continues it. The repo's own README explains why: Stolen Sword's concept was carried forward into a separate, paid mobile release, Wild Wind, on iOS and Android. That's a perfectly reasonable thing for an indie developer to do with a successful jam idea, but it does mean the free, open-source game under review here plays like an extended proof-of-concept for a product this review has no way to cover. Dying only costs you the current wave — two hits and you restart it from its start, no larger penalty — which keeps the combat approachable but also fairly low-stakes, and combined with the short runtime there's little tension left once the gesture clicks.

Verdict

Stolen Sword earns its shelf spot on the strength of one thing: a control scheme that is genuinely, verifiably built for a single thumb, with a source-level guarantee that mouse and touch behave identically. It's satisfying to pull off once you learn it works backward from instinct, and the four stages spend their tiny file-size budget on real structural variety instead of repetition. But it's a fifteen-minute experience that ends unresolved by design, with nothing built in to teach its own most important idea in words, and 72 GitHub stars plus 17 forks — a solid showing for a js13k entry — reflects a clever jam curiosity more than a game with lasting pull. Recommended for the ten minutes it takes to feel the mechanic work; don't expect much beyond that.