Girder Span skill runs: ledge perfects on a rising beam
What the first skill runs second looks like
Girder Span opens on a construction balcony with a single instruction: hold to extend. That is the anchor. You press, a steel plank grows from under your boots, you release, the plank drops into a bridge, and the worker runs. Miss the far ledge and the run ends in a slow fall. Nail the gold edge of the platform and the desk logs a perfect.
Under the cantilever method this is the cleanest skill runs loop of the three. The verb never splits. There is no second stick, no inventory, no map. Reach is measured in spans and floors; deflection arrives later as wind that drifts the falling plank by a few percent of its length.
Load test: Anchor
On a cold start the hold prompt is readable within the first second. Mouse, touch and Space all grow the girder. We confirmed that walking ignores input, which prevents accidental re-holds mid-span. On a 390px-wide viewport the prompt still clears the Close control after the HUD padding shift.
Load test: Reach
A first-run player typically dies between spans three and eight. Protocol bots hold longer — twenty to forty seconds, four to six floors — because they aim for centre rather than the perfect strip. That is fine for skill runs scoring: the fail state is always a fall, never a soft reset into a menu.
Load test: Deflection
Wind is the main bend. After roughly eighteen seconds a gust banner appears and the dropped plank shifts sideways. Gold platforms pay double but widen the gap. Night shift and crane swing events change the lighting and timing without introducing a new control scheme. That is the deflection we want: one lever, same verb.
Load test: Load
Three protocol passes on desktop produced no input conflicts. The narrow pass showed the camera framing the current balcony and the next ledge without clipping the Close link. Combo audio pitches up on perfect streaks; we treat that as feedback, not as a second system to learn.
Load test: Release
The review only ships because Close returned from every state. Play opens at /play/girder-span/. The ranking card uses a landscape preview cropped from a live play frame — not a portrait phone shot.
Strengths we logged
- Single hold-release verb that teaches itself
- Perfect window that rewards eye measurement without a tutorial slide
- Wind and gold as optional risk, not mandatory modes
- Readable fail with slow-motion fall
Limits we refuse to dress up
- No public store rating or download count exists — we leave those cells blank
- Perfects are strict; centre-aim bots will under-report the skill ceiling
- Theme is construction Stick-Hero; if you want ice or flight, look at ranks 2–3
How a protocol pass is timed
We start the clock on the first hold, not on the menu tap. A pass ends on the over screen. Between passes we reset skins to the default hard hat so cosmetics do not leak into the log. Missions and star thresholds are noted but do not change rank — the cantilever ranks clarity, not completionist pressure.
Gold platforms are logged as optional deflection. If a pass never takes a gold span, we still accept it, but the review will say so. Crane swing is treated as a timing gate on the walk, not as a new verb. Night shift is a palette and scoring buff; controls stay identical.
Why this skill runs build sits above the other two
Girder Span teaches the perfect window without a glossary. Gust Path needs you to trust inertia. Crack Chase needs you to trust the lead band. On a desk that publishes skill runs for first-time visitors, the hold-release beam is the one that fails most honestly. That is the only reason it is rank one.
Session notes from 5 November 2026
Pass A: death on span 5, two perfects, no wind yet. Pass B: floor 3, gust survived, one gold. Pass C: crane clip on span 11. Narrow pass: same loop, Close never covered. These notes are the evidence behind the side facts card — not a marketing paragraph.
Verdict on the cantilever
Girder Span takes rank one because skill runs stay legible when the only question is “did the beam hit the ledge?” Best for ledge perfects. Checked 5 November 2026.