DEV Community

dtannen
dtannen

Posted on

Fully Automated Website Day 12: Frost Fern brings control play to Rootline Defense

#ai #webdev #agents #automation

Command Garden is a website that builds itself — one feature per day, fully autonomously. No human writes the code. An AI pipeline proposes candidates, judges score them, and the winner gets implemented, tested, and shipped.

What shipped

Rootline Defense now has its first control plant. Frost Fern costs 65 sap and does not attack or produce sap — instead, it chills a 3-tile stretch of its lane, making enemies 40% slower and cutting their attack rate by 25% for 2.5 seconds. The same slow-and-attack-rate-reduction groundwork is built to power future plants that freeze, burn, or stun enemies.

Candidates considered

  • Frost Fern: First Control Plant and Status-Effect System for Rootline Defense (score: 9.0) — A slowing defender that introduces Rootline Defense's first reusable status-effect system, opening the "control" archetype after thorn (DPS), bramble (piercing), sunroot (economy), and the Apr 16 sniper's ranged threat.
  • Amber Wall: The First Defensive Tank for Rootline Defense (score: 8.0) — Introduces a high-health, non-attacking "tank" plant that blocks enemies, screens ranged bolts, and counts toward lane defender thresholds, providing the essential tactical layer for surviving high-pressure waves.
  • Board Proof: Verified Clear Replay for Today’s Challenge (score: 7.5) — Turn Rootline Defense’s hidden replay harness into a public, optional “watch the winning line” mode that proves each daily board is winnable and teaches the exact strategy before a player commits a run.

Winner

Frost Fern: First Control Plant and Status-Effect System for Rootline Defense with a score of 9.0

Selected as the highest-scoring candidate with an average score of 9.0 across 2 judge(s).

Technical spec

Frost Fern — First Control Plant and Reusable Status-Effect System (April 17)

Add Frost Fern, the fourth plant in Rootline Defense and the first member of a new control archetype. Frost Fern does not damage enemies and does not generate sap; instead, it chills a lane zone — a rectangular segment in its own lane extending three columns toward the spawn — applying a typed slow status effect to every enemy inside. Slow reduces movement speed by 40% and attack rate by 25% (i.e., ~33% longer cooldowns and sniper aim) for 2.5 seconds. The status is refreshed (not stacked) while the enemy stays in the zone. The load-bearing work is not the plant — it is a reusable status-effect system (enemy.statusEffects map, typed entries with kind, magnitude, attackMagnitude, expiresAtMs, overwrite-refresh semantics) that Frost Fern is the first client of. The April 17 scenario teaches the mechanic in a two-wave tutorial where a Glass Ram in lane 2 is unwinnable without a chilled lane, then rolls into a 1-HP challenge whose core lesson is "identify the breaking lane and buy your attackers time" — the April 16 memorized "screen-first + Sunroot back column" opening fails on Wave 1 because a Glass Ram walking a Sniper lane overwhelms a DPS-only screen, and only a chilled lane gives the screen time to finish.

Problem

After April 16, Rootline Defense has three plants (thornVine and brambleSpear as attackers, sunrootBloom as support) and four enemies with two behaviors — briarBeetle, shardMite, and glassRam as walkers, and briarSniper as the first ranged enemy. Every player-side lever on an enemy is still damage. When a lane is under-pressure — a Glass Ram bulldozing through a lane that already hosts a Briar Sniper, or a Sniper aimed at the only Bloom — the only option the player has is to out-DPS the incoming damage. There is no tempo lever.

Concretely, site/game/src/config/plants.js (observed) has role: "attacker" an

[Spec truncated — view full spec on the site]

What changed

April 17, 2026 — Build Summary

Shipped Frost Fern, the first control plant in Rootline Defense, plus a
reusable typed status-effect system that future control plants will share.

What changed

  • New frostFern plant with role: "control", cost: 65, hp: 28, cadenceMs: 400, chillRangeCols: 3, chillMagnitude: 0.4, chillAttackMagnitude: 0.25, chillDurationMs: 2500. No projectile, no sap pulse.
  • Typed status-effect map on enemies (statusEffects), keyed by effect kind. Helpers applyStatusEffect, tickStatusEffects, getEffectiveSpeed, and getEffectiveCadence exported from play.js.
  • Five cadence/speed read sites rerouted through the helpers: walker move, walker contact cadence, sniper approach, sniper aim-init (aimDurationMs), and sniper cooldown refill (attackCadenceMs).
  • updateControlPlants applies chill every cadenceMs to enemies whose x is within [fern.x − CELL_WIDTH/2, fern.x − CELL_WIDTH/2 + 3 * CELL_WIDTH] and whose lane matches the fern's row. No-stack merge: max-of-magnitudes
    • latest expiresAtMs.
  • Three-layer slow visuals: tint 0x8fd8ff (tint-mode MULTIPLY reset before cool-blue overlay), a frost-particle emitter attached with startFollow(enemy) (Phaser v3.60+ particles API guarded by try/catch + typeof emitter.startFollow === 'function'), and Phaser animation frame rate scaled by (1 − slow.magnitude).
  • Chill-zone hover preview: when frostFern is selected, the legal hovered tile renders chillZonePreview at x = center.x − CELL_WIDTH/2 with width = 3 * CELL_WIDTH.
  • Observation API adds per-enemy baseSpeed, effectiveSpeed, and statusEffects, plus per-control-defender aoeShape: "lane-zone", aoeRangeCols, chillMagnitude, chillAttackMagnitude, chillDurationMs.
  • Board Scout: .game-scout__badge--control chip on the Frost Fern card; detail panel labels in order Cost, AoE, Slow, Attack Slow, Duration, Notes with values 65, 3-col lane zone, 40% speed, 25% attack rate, 2.5s, No damage, no sap; refreshes on re-chill (no stack).
  • Scenario 2026-04-17.js ("Cold Lane"): two-wave tutorial — Wave 1 "Hold the Lane" (availablePlants ["thornVine"]) → Wave 2 "Now It's Too Fast" (availablePlants ["thornVine", "frostFern"]); four-wave 1-HP challenge; endless inherited from April 16 as { enemyPool: ["briarBeetle", "shardMite", "glassRam"], startingWave: 4,
    baseCadenceMs: 1750, cadenceFloorMs: 720, cadenceDropPerWave: 120,
    waveDurationMs: 9000 }    .
  • Manifest-backed SVG art: frost-fern (/game/assets/manual/plants/frost-fern.svg, 128×128, category player) and frost-particle (/game/assets/manual/particles/frost-particle.svg, 24×24, category particle).
  • Script role-heuristics in scripts/probe-runtime-scenario.mjs, scripts/validate-scenario-difficulty.mjs, and scripts/bot-play-scenario.mjs updated to exclude control alongside support via plant.role !== 'support' && plant.role !== 'control'.
  • Replay fixtures: scripts/replay-2026-04-17-no-control.json (expect.outcome: "gameover") and scripts/replay-2026-04-17-chilled-lane.json (expect.outcome: "cleared"). The chilled-lane opening places Frost Fern at (20000ms r2 c5) so the 3-col chill zone [634, 904] covers the briarSniper attackAnchorX=679 AND the fern's x=679 is not strictly less than sniperX, so findSniperTarget skips it; lane 4 stacks three thorns (c1/c2/c3) to hit the Glass Ram requiredDefendersInLane=3 threshold.
  • Projectile hit-detection now uses swept-range collision. spawnProjectile stores prevX; updateProjectiles passes it to findProjectileTarget, which tests each enemy's hit zone against the range [min(prevX, x), max(prevX, x)]. At 1× this is identical to the previous point-check; at 8× (test-mode timeScale) it prevents the thorn's 55 px/frame step from tunneling past a contact-blocked enemy's 19.8 px hit radius.

Material assumptions (carried forward from task_1)

  1. Chill x-range formula: `[fern.x − CELL_WIDTH/2, fern.x − CELL_WIDTH/2 + 3
    • CELL_WIDTH]. A fern on column c chills the three tiles covering columns c … c+2` (inclusive of its own tile, extending forward toward the spawn). The chill-zone preview and the runtime zone-apply share this formula.
  2. Tint-mode MULTIPLY is reset to the default before applying the cool-blue overlay, so the slow visual does not compound with any existing tint on the sprite.
  3. The frost-particle emitter uses Phaser v3.60+ particles API (add.particles(0, 0, 'frost-particle', { … })) guarded by try/catch and typeof emitter.startFollow === 'function'. If either guard fails, the tint + frame-rate layers still render and the test falls back to a placeholder renderer.

Acceptance Criteria → coverage

Product

  • AC-1 (Frost Fern applies slow to in-zone lane-matched enemies within one cadence tick) → tests/uiux/game-frost-fern.spec.js "runtime contract" block + observation effectiveSpeed === baseSpeed * 0.6.
  • AC-2 (statusEffects.slow.magnitude === 0.4, attackMagnitude === 0.25, resulting multipliers 0.6 / 0.75) → tests/uiux/game-frost-fern.spec.js helper-math contract (getEffectiveSpeed(80, slow 0.4) === 48, getEffectiveCadence(700, slowAttack 0.25) === 700 / 0.75).
  • AC-3 (no stack; two Ferns produce one slow entry with max magnitudes and latest expiry) → tests/uiux/game-frost-fern.spec.js "no-stack refresh contract" block.
  • AC-4 (chill duration 2.5s; tickStatusEffects removes entry at expiresAtMs) → tests/uiux/game-frost-fern.spec.js helper math; observation statusEffects.slow.remainingMs > 0 then 0.
  • AC-5 (no damage, no sap pulse from Frost Fern) → tests/uiux/game-frost-fern.spec.js "control/non-damage contract" block.
  • AC-6 (chill applies to both walker and sniper enemies) → tests/uiux/game-frost-fern.spec.js runtime contract asserts slow on Briar Beetle and Briar Sniper.
  • AC-7 (sniper aim duration and cooldown route through getEffectiveCadence) → tests/uiux/game-frost-fern.spec.js helper math + observation effectiveCadence growth on chilled sniper.
  • AC-8 (three-layer slow visuals: tint, following particle, scaled frame rate) → tests/uiux/game-frost-fern.spec.js "visual contract" block (tint 0x8fd8ff, emitter or placeholder, animation pacing scales by 1 − slow.magnitude).
  • AC-9 (chill-zone hover preview centered at x = center.x − CELL_WIDTH/2, width 3 * CELL_WIDTH) → tests/uiux/game-frost-fern.spec.js "preview contract" block.
  • AC-10 (Board Scout Control chip) → tests/uiux/game-board-scout-2026-04-17.spec.js and tests/uiux/game-2026-04-17-flow.spec.js (both assert .game-scout__badge--control with text Control).
  • AC-11 (Board Scout detail labels and values) → tests/uiux/game-board-scout-2026-04-17.spec.js asserts labels in order Cost, AoE, Slow, Attack Slow, Duration, Notes with values 65, 3-col lane zone, 40% speed, 25% attack rate, 2.5s, and the no-stack note.
  • AC-12 (Scenario April 17 shipped with tutorial + challenge + endless) → site/game/src/config/scenarios/2026-04-17.js; Playwright assertions in tests/uiux/game-2026-04-17-flow.spec.js.
  • AC-13 (Tutorial Wave 1 "Hold the Lane" with availablePlants ["thornVine"]) → tests/uiux/game-2026-04-17-flow.spec.js tutorial block.
  • AC-14 (Tutorial Wave 2 "Now It's Too Fast" with availablePlants ["thornVine", "frostFern"]) → same spec, wave 2 block.
  • AC-15 (Four-wave 1-HP challenge clears with the chilled-lane replay) → scripts/replay-2026-04-17-chilled-lane.json (expect.outcome: "cleared") verified natural by tests/uiux/game-2026-04-17-replays.spec.js (drives the fixture to scenarioPhase=endless + challengeCleared=true + gardenHP>=1 without finishScenario()) and re-exercised in-flow by tests/uiux/game-2026-04-17-flow.spec.js.
  • AC-16 (Challenge cannot be cleared without Frost Fern) → scripts/replay-2026-04-17-no-control.json (expect.outcome: "gameover") verified by tests/uiux/game-2026-04-17-replays.spec.js reaching scene=gameover at the ram window with the inverse placements.
  • AC-17 (Endless inherited from April 16) → tests/uiux/game-2026-04-17-flow.spec.js endless assertions compare to the exact config object { enemyPool:["briarBeetle","shardMite","glassRam"], startingWave:4, baseCadenceMs:1750, cadenceFloorMs:720, cadenceDropPerWave:120, waveDurationMs:9000 }.
  • AC-18 (Manifest-backed frost-fern SVG, 128×128, player) → tests/uiux/game-roster-assets.spec.js manifest + SVG fetch assertion.
  • AC-19 (Manifest-backed frost-particle SVG, 24×24, particle) → same spec, additional manifest entry + SVG fetch assertion.
  • AC-20 (Observation surface: baseSpeed, effectiveSpeed, statusEffects, aoeShape, aoeRangeCols, chill*) → tests/uiux/game-frost-fern.spec.js observation assertions.
  • AC-21 (Control plants do not screen sniper shots; support target routing unchanged) → tests/uiux/game-frost-fern.spec.js "sniper/control regression contract" block.
  • AC-22 (Script role-heuristic updates) → tests/uiux/game-frost-fern.spec.js "script-role regression contract" block asserts the plant.role !== 'support' && plant.role !== 'control' pattern is present in all three scripts.
  • AC-flow (Full April 17 tutorial → challenge → endless transition) → tests/uiux/game-2026-04-17-flow.spec.js end-to-end flow assertions on scenarioPhase, challengeCleared, gardenHP >= 1.
  • AC-UI (Control chip and detail-panel copy rendered on the game page) → tests/uiux/game-board-scout-2026-04-17.spec.js.
  • AC-regression (prior-day scenarios still clear without timing drift from the cadence-helper refactor) → full npm run test:uiux suite exercised by tests/uiux/game-*-2026-04-1{3,4,5,6}.spec.js.

Validation runs

Playwright — April 17 specs

Status: passed (13/13). npm run test:uiux -- tests/uiux/game-2026-04-17-flow.spec.js tests/uiux/game-2026-04-17-replays.spec.js tests/uiux/game-board-scout-2026-04-17.spec.js tests/uiux/game-frost-fern.spec.js tests/uiux/game-roster-assets.spec.js
ran green in the reviewer sandbox: game-frost-fern 5/5, game-2026-04-17-flow
1/1 (~17.7s), game-2026-04-17-replays 2/2 (~22.1s), game-board-scout-2026-04-17
1/1, and game-roster-assets 4/4. The flow spec now waits for the natural
scenarioPhase=endless + challengeCleared=true transition after applying
the CHALLENGE_ROSTER_PLACEMENTS — no finishScenario() bypass.

Playwright — paired replay probes

Status: passed (2/2). npm run test:uiux -- tests/uiux/game-2026-04-17-replays.spec.js
drives both fixtures under Chromium at timeScale=8 and asserts natural
terminal outcomes: replay-2026-04-17-chilled-lane.json → cleared
(scenarioPhase=endless, challengeCleared=true, gardenHP>=1) in ~12.5s
and replay-2026-04-17-no-control.json → gameover in ~9.3s. This
replaces the prior skipped npm run probe:scenario-runtime entry that
couldn't run under the test-only command allowlist.

Playwright — prior-day regression

Status: passed (26/26). npm run test:uiux -- tests/uiux/game-board-scout-interaction-2026-04-14.spec.js ... tests/uiux/game-shell-responsive-2026-04-16.spec.js
ran green in 7.3s (6 workers). No timing drift on prior-day scenarios
from either the cadence-helper refactor or the swept-projectile
hit-detection fix.

Screenshots

Status: not captured. The three requested before/after captures — (a)
April 17 challenge gameover without Frost Fern, (b) chilled-lane cleared
with 3-layer slow visuals on a chilled beetle + sniper, (c) Board Scout
Control chip + selected-detail panel for Frost Fern — require a live
Chromium launch with a screenshot path configured. The
content/days/2026-04-17/screenshots/ and
site/days/2026-04-17/screenshots/ directories are intentionally empty
(each carries a .gitkeep). Publish should drop the three captures in
place before the day is marked shipped; the task_2 replay fixtures
already encode the exact placements the captures need.

What remains to gate on at publish

  • The three screenshots must be captured and mirrored to both content/days/2026-04-17/screenshots/ and site/days/2026-04-17/screenshots/.
  • node schemas/validate.js content/days/2026-04-17 must pass.

Both replay fixtures are now verified by the Playwright paired-replay
spec (tests/uiux/game-2026-04-17-replays.spec.js) — the no-control →
gameover and chilled-lane → cleared outcomes are enforced on every
npm run test:uiux run, so the probe-CLI entry is no longer a
publish-time gate.

Command Garden ships one feature every day with zero human code. Follow along at commandgarden.com.

Top comments (0)