Second Mobile Feedback Round

This is the sanitized review ledger for direct-message and observed playtest feedback received after the 0.8.1-beta.1 release on 2026-08-06. It contains no account identifiers, device identifiers, IP addresses, or private-message metadata.

Batch Overview

F12 — First Ten Minutes Feel Like Hurried Work

• Tester: 小南 · direct observed session
• Report: After more than ten minutes, the tester stopped because play felt like going to work and required constant attention.
• Corroboration: 丁卷儿 independently described the prototype as closer to a Kairosoft management game than an idle game because plants grow quickly and create repeated clicks. This reinforces F03 rather than replacing it.
• Decision: Treat the first ten minutes as a dedicated design test. Do not change all timing constants inside a navigation patch. First remove search and scrolling work, then compare action count, unattended tolerance, and automation discovery on a fresh save.
• Success signal: A new player can leave the garden alone for a short interval, return, understand what changed, and continue without feeling punished or losing the thread.
• 2026-08-09 retest (simulated, not yet human-verified): With navigation already fixed and the F30 bloom-ending warning already shipped, a simulated-player harness measured plants lost when checking in only every 90/120/180 seconds. Losses were real (8/14/20 over 15 minutes) and split between bloom-timeout and a separate thirst-stress decline mechanic. Fixed with a targeted 1.8× bloom-duration increase and a halved severe-dry/wet stress-accumulation rate; losses at 90s check-ins dropped to 0, at 120s to 3. The 7-9 actions/minute workload at short check-in intervals was measured but deliberately not changed — see decision-log.md and dev-log 2026-08-09 for the full numbers. This addresses the "unattended tolerance" half of the success signal; whether the moment-to-moment feel still reads as "hurried work" during active play is still an open, human-only question.

F13 — Mobile Seed Choice Appears To Be Mint Only

• Tester: Reported by every phone tester in this round
• Report: Players asked whether mint was the only plant. The compact current-seed control did not communicate that it opened a list or that another starter seed was already available.
• Decision: Make the control state the number of currently plantable species and use an explicit change affordance. The closed state must communicate choice without requiring prior knowledge of a bottom sheet.
• Follow-up: Test a fresh save before and after planting mint. The presence of calendula must be discoverable without scrolling below the greenhouse.

F14 — Expected Pot-First Planting

• Tester: 丁卷儿 · phone
• Report: The first instinct was to tap an empty pot and then choose what to plant. The existing seed-first flow only became apparent after repeatedly planting mint.
• Decision: Support both mental models. Keep the global selected seed for fast repeated planting, and add an empty-pot route that opens the seed choices for that specific pot and plants immediately after selection.
• Follow-up: The targeted picker must name the pot, preserve soil preview information, and cancel safely without planting.

F15 — Irrigation Is Hard To Relocate

• Tester: 丁卷儿 · phone
• Report: The first trial dripper appeared to work on one pot and then seemed to disappear. Later irrigation purchases were eventually found under equipment, but the route was not memorable.
• Decision: When a selected pot has no irrigation, expose a local action that either installs owned equipment or opens the equipment catalog. Keep the catalog as the purchasing home; do not duplicate its full contents inside every pot.
• Follow-up: Test both inventory states: one owned dripper and zero owned drippers after tools unlock.

F16 — Soil State Is Not Legible On The Pot

• Tester: 丁卷儿 · phone
• Report: Soil adjustment felt indirect. The tester suggested using pot color, such as blue framing for acidic soil and green for ordinary soil.
• Decision: Give every pot a persistent soil-profile accent and a compact profile label. Blue denotes acidic, green ordinary, and rose denotes alkaline. Color is redundant with text so the state does not depend on color vision alone.
• Follow-up: Verify empty and occupied pots, selected states, and hydrangea color previews at phone width.

F17 — Goals And Tools Remain Outside Attention

• Tester: 大花 · phone-oriented layout feedback
• Report: Orders that earn coins and equipment remained too far below the initial viewport; growing seed lists compounded the long-page feeling.
• Corroboration: This repeats F06 even after workspaces were introduced, which suggests the remaining issue is workspace discoverability rather than raw document order alone.
• Decision: Keep separate mobile workspaces, but make the top greenhouse controls explicitly name 种子选择 and 设备与自动化. Preserve the goal strip as a direct route to orders.
• Follow-up: Fresh-save test whether a player can name where seeds, orders, and irrigation live after one exposure.

F18 — Pot Alerts May Still Be Too Quiet

• Tester: 大花
• Report: Abnormal pot states required close inspection.
• Version uncertainty: The comment may refer to the previous-night build, before the latest drought treatment was online.
• Decision: Do not immediately add another alarm layer. Retest the current live build and separately check dry, wet, aphid, declined, and harvest-ready states. Promote only states that still fail recognition.
• 2026-08-09 code audit: Checked dry, wet, aphid, declined, bloom-ending, and harvest-ready states individually against their render code rather than assuming the round's other fixes already covered this. Found two states that were still under-signaled by design, not just by perception: severe overwatering had no equivalent to the severe-dry crisis stamp despite an equal stress penalty, and aphid infestation used the same neutral chip styling as a friendly visitor with no color distinction at all. Both promoted to match the existing severe-dry/dry-alert treatment. Dry, declined, bloom-ending, and harvest-ready were already adequately signaled and were not changed further. Still needs a human pass to confirm these read correctly in actual play, not just in code.

Round Decision

Implement F13 through F17 as a mobile discovery and action-locality candidate. Keep F12 as the next structured balance test and F18 as a version-sensitive visual retest. The candidate should reduce navigation labor without claiming that the idle-versus-management identity question is solved.

F19-F22 Follow-up

• F19: Keep the full soil comparison surface on desktop, but remove it from the mobile document flow. Empty pots open a targeted soil drawer where products can be bought and applied without searching for a separate laboratory.
• F20: Every order ingredient shows current stock, required stock, and an explicit 还缺 N state. The top order reminder repeats exact deficits.
• F21: During Cup cultivation, show a quiet status strip that can pause only the main greenhouse. Event plants continue to grow; global pause remains available for stopping everything.
• F22: Rename the mobile destinations 种子库 and 温室商店, add compact identifying marks, and keep pot-local shortcuts as routes into those homes rather than duplicate systems.
• F23: Remove the permanent 查看自动浇水 action from every selected pot. Keep immediate local installation only when a dripper is already owned. Signal owned equipment or an affordable permanent upgrade on the single 温室商店 entry without treating ordinary consumables as a permanent alert.
• F24: The first soil-drawer release placed its trigger inside a summary that mobile compression deliberately hid. Empty pots now expose a dedicated compact soil action. Occupied pots make their soil-state label interactive, allowing preview and advance purchase while clearly withholding application until harvest.
• F25: Mobile pot selection now opens a compact action dock by default. The current action remains available, while calculations and automation explanations move behind an explicit detail toggle.
• F26: Pause and resume become a persistent item in the mobile bottom navigation. Speed options remain preference controls; they are not counted as a solution to required action volume or attention switching.
• F27: Empty-bed actions are derived from actual available inventory, not only the last selected seed. One remaining species becomes a direct planting action; multiple species open explicit selection; zero species states the shortage.
• F28: Soil uses a complete two-pixel profile border and a filled profile label. Selection uses an inner focus ring so it does not erase the soil color.
• F29: Pause remains persistent but uses its own tinted control block and a strong divider from workspace navigation. Its active resume state changes color as well as icon and text.

Third Round: F30–F33

This batch was collected directly by the creator after the 0.8.4-beta.2 release on 2026-08-07, and handed to a new AI session (Claude) after the original session ran out of budget. Two reports came from 小明 through the in-game feedback form; one is a directly observed session with 丹丹; F33 is a direct creator report found while reviewing this same round of fixes.

F30 — Bloom Decline Has No Warning; Water Alerts Too Subtle; Confusing Copy

• Tester: 小明 · category 想法 · 温室与花盆操作 · v0.8.4-beta.2 · viewport 874×358
• Report (verbatim, sanitized of tester id): "花多开始衰败没有提示，然后一下就死了体验很差。另外浇水要进行操作的提示不够明显，红色还不够明显，缺水不够明显。给的时间窗口有点短。有点装死这个说法莫名其妙"
• Root cause: getStage() swapped the stage label to 花期将尽 at 70% of bloom age with zero visual signal (no color, no badge, no card treatment); the top care summary counted it identically to ordinary "可收获" the whole time. Separately, the mild "偏干" state carried only a thin 4px left border in a muted red (--danger: #a8473d), and the severe-dry crisis stamp's entrance animation only played for one 1.35s tick before going static.
• Decision: Add a distinct pulsing amber card treatment plus a dedicated 花期将尽 N count and top-priority routing for bloom-ending plots. Strengthen dry-state visuals with a more saturated --warning-red, a full card border instead of a thin edge, a solid-filled water chip, and a 3-tick alert window. Reword "生命体征：有点装死" to "严重缺水，需要立刻浇水".
• Follow-up: Retest whether the new warning is noticed before a plant declines, not just after. bloomDuration itself was deliberately left unchanged this round (see decision-log 2026-08-09); if testers still call the window too short after the warning ships, that becomes the next numeric-tuning candidate.

F31 — Feedback Modal Obstruction and No Pot-Deselect Route

• Tester: 小明 · category 故障 · 温室与花盆操作 · v0.8.4-beta.2 · viewport 874×358 (captured after rotating to landscape as a workaround)
• Report (verbatim, sanitized of tester id): "目前ui有点问题，如果不巧点击到花盆激活了操作或详情页面，会遮挡反馈界面的下部分内容。而且点击空白处无法取消花盆选择。另外最下面的切换界面bar遮住了竖屏情况下的发送键，必须横过来才能发送"
• Root cause: .feedback-modal was z-index: 30, lower than the mobile pot-detail dock (35/60) and the persistent bottom workspace nav (40), so those overlays could sit on top of the feedback sheet's lower half and submit button in narrow portrait widths. Separately, #contextBar never exposed a close control or an outside-click handler, so a selected pot could only be replaced by selecting another one.
• Decision: Raise the feedback modal to the top of the stacking order (z-index: 200, above every game overlay including the wheel-prize reveal). Add an explicit × close button on the pot context bar, a click-on-blank-grid-area handler, and Escape-to-deselect.
• Follow-up: Retest the feedback form specifically at narrow portrait widths (~360–400px) with a pot selected underneath, to confirm the submit button is always reachable without rotating.

F32 — Observed Session: Sudden Wilt Breaks Trust, Play Feels Like Obligation

• Tester: 丹丹 · directly observed session, not a form submission
• Report: Scrolled up briefly to look at something else; came back to find a flower had gone from blooming to withered with no warning in between. Described the game as feeling more like "attention is mandatory or you get punished" than something that hands out small rewards for coming back. Corroborates 小南's earlier "feels like going to work" comment (F12) and 丁卷儿's Kairosoft-management comparison.
• Decision: Same fix as F30 — this is the same missing warning tier, observed from the player side rather than reported through the form. Treat F12 as still open (structured pacing retest) rather than closed by this UI fix; the warning system removes the surprise of decline, but does not by itself prove the pacing itself feels generous rather than demanding.

F33 — Lucky Wheel Prize Reveal Flashes Repeatedly

• Tester: Creator · direct report during this session
• Report: "抽到奖品的闪动页面很奇怪，不连贯，一段一段地闪，很像bug，或者是那种反复提醒得到了同一个奖品的效果，而不是highlight当前奖品。"
• Root cause: The reveal card's re-render guard only covered the spin phase, not the reveal-pending phase. Every background game tick fully rebuilt the reveal overlay's DOM while it sat on screen, restarting its entrance/confetti animation each time.
• Decision: Extend presentationAlreadyMounted in renderEvent() to also hold while state.event.wheel.revealPending is true. Verified with a standalone Node harness counting innerHTML writes against both the old and fixed code.
• Follow-up: Retest a full wheel draw live, ideally leaving the reveal on screen for several seconds before acknowledging, to confirm the card now holds still.
• Follow-up: Next structured F12 retest should specifically check: does a player who steps away for 1–2 minutes come back to an understandable, recoverable state now that the warning exists, or does the underlying bloom duration still need lengthening?

Creator Audit: F34-F35

F34 — Desktop Soil Action Opens a Hidden Destination

• Report: On a large screen, clicking a pot's soil-profile action produces no visible response.
• Root cause: Desktop workspace tabs hide the standalone soil laboratory, but the click handler still only selected that hidden element and attempted to scroll to it.
• Decision: Treat soil configuration as one targeted panel at every width. The same pot-local action now opens the soil panel directly, while phone retains its bottom-sheet presentation and larger screens receive a centered dialog.

F35 — Desktop Context Bar Covers Pots

• Root cause: The selected-pot context bar retained its old bottom-sticky positioning after the desktop garden became a bounded workspace, so it could float over the final row.
• Decision: Keep sticky/fixed action access on phones, where reach matters, but return the desktop context bar to normal document flow.