Create an index.html that includes p5.min.js, and implement an L-system tree with branches drawn recursively, with sway (swaying) added via Perlin-noise wind applied to branch angles. Self-contained.

Design: L-system. Axiom "F", rules F -> "FF+[+F-F-F]-[-F+F+F]" or a simpler one. Grow progressively (animate growth). Swaying by adding a noise-based wind offset to the angle at each depth, scaled by depth.

Implementation: generate the string, parse into a branch structure (tree of segments), animate growth with progress 0..1, draw recursively using turtle. Wind: at each branch, angle += wind(t) * depthFactor.

Use p5 instance mode or global mode. Global mode is fine.

Drawing: precompute a tree of nodes: each segment has len, baseAngle, children, depth, growthStart. Growth animation: overall progress t; each branch grows in order with depth. Simpler: growth is progress per depth level — each branch is drawn only when global growth reaches its depth, and length scales.

Leaves at terminal branches. A ground gradient, sky, maybe a subtle glow. Keep it reasonably nice.

Add controls: click/space to replant with a new seed. HUD text.

Let me check if it works — is node available? Let's do a quick sanity check on JS syntax with node (p5 doesn't exist, but parse check is possible). Can I use headless p5? More simply: extract the script and syntax-check it with node --check.

One concern: the L-system strings contain "X", which doesn't draw anything, but "F" continues along the current path — in my parser, each F creates a new child and then descends into it, forming a chain (correct). "[" pushes, "]" pops. Good.

One issue: leafTipX/Y is returning 0, so the falling leaves spawn from the base of the trunk — trivial issue but visible. Improvement idea: record the leaf's world position (relative to the root) when collecting leaves. collectLeaves is called with the root's orientation (angles relative to up, -90° rotation applied at draw time). Positions are computed in tree space where F draws in the angle 0 direction. At draw time, the space is rotated -90°. So leaf world coords = (x, y) rotated by -90°: screen dx = y? With -HALF_PI rotation: (x,y) -> (y*?, ...). Rotation by -90°: (x, y) → (x cosθ − y sinθ, x sinθ + y cosθ), with θ=-π/2: (y·1? cos=0, sin=-1 → (−y·(−1)? Let me just compute: newX = x·0 − y·(−1) = y; newY = x·(−1) + y·0 = −x. So screen offset = (y, −x). Store lf.x, lf.y in tree space and transform them.

Also, the leaves array gets pushed during collectLeaves, but plantSeed resets the leaves afterward? Order: tree = buildTree; leaves = []; collectLeaves — good.

But collectLeaves ignores the branch angle (wind/rotation) when computing positions — the angle passed in stays 0 forever. Fix: collectLeaves should accumulate ang + b.angle. But b.angle is set during buildTree; collectLeaves runs after that. Currently ang doesn't change. Fix this to compute approximate static positions (ignoring wind).

Also, regarding the wind accumulation in drawBranch: inheritedWind is being passed, but ang includes inheritedWind, and it's further accumulated by passing `inheritedWind * 0.4 + w * 0.5` to children — no problem.

One more check: in buildTree, the children of root are each F from the root... axiom "X" doesn't produce F at the root level, but the rules expand X into a string containing F, so it's fine. In the case of axiom "F", the root gets one child chain. The trunk: a chain of F segments, and since the length decays with depth via lenShrink^(depth-1), the trunk segments also gradually get shorter — acceptable, but the trunk might look segmented. Since depth 1 uses lenShrink^0 = baseLen, the first several are at full length. OK.

Recheck the syntax.

If p5 is available in node, maybe I can do a simple runtime smoke test? Not installed. I could also run it headless with jsdom, but that's heavy. Syntax check + logic review is probably sufficient. Let me quickly check the final file to verify correctness.