Turnip Prophet - Animal Crossing Turnip Price Predictor

Predicts Animal Crossing: New Horizons turnip prices using the game's actual algorithm.

⚠️ IMPORTANT: Always Read Memory First!

Before doing ANYTHING, read the weekly data file:

memory/turnip-week.json

When the user gives a new price, update memory/turnip-week.json immediately with the new value before running the prediction.

Weekly Data Format (memory/turnip-week.json)

{
  "week_start": "2026-02-15",
  "buy_price": 96,
  "previous_pattern": 1,
  "prices": [84, 81, 78, null, null, null, null, null, null, null, null, null],
  "labels": ["Mon AM", "Mon PM", "Tue AM", "Tue PM", "Wed AM", "Wed PM", "Thu AM", "Thu PM", "Fri AM", "Fri PM", "Sat AM", "Sat PM"]
}

On Sundays, create a fresh file with the new buy price and reset prices to all nulls.

Triggers

Activate when user mentions:

• "turnip prices" or "turnip price"
• "ACNH turnips" or "Animal Crossing turnips"
• "stalk market"
• "turnip prophet" or "turnip prediction"
• "bell profit" with context of turnips
• Any question about when to sell turnips in Animal Crossing

Privacy & Data Storage

This skill stores local configuration data only on your machine in memory/turnip-config.json. No data is sent to external servers.

What gets stored (if you enable reminders):

• Channel name (telegram/whatsapp/discord/signal)
• Your user ID on that channel
• OpenClaw binary path
• Setup timestamp

Where: Stored locally in the skill's memory directory on your OpenClaw instance. Not shared, not uploaded, not visible to anyone else.

How to disable/reset:

rm ~/.openclaw/workspace/skills/turnip-prophet/memory/turnip-config.json

Removing cron entries:

# Edit crontab and remove the turnip-prophet lines
crontab -e

Important: You must explicitly confirm the setup flow before any config is saved. Declining the setup offer means nothing is stored.

Credentials & Permissions

The skill itself does not request or store any API keys or credentials.

However, if you enable cron reminders, the installed cron jobs will:

• Use your local openclaw CLI binary
• Send messages using your existing OpenClaw configuration and credentials
• Operate with whatever permissions your OpenClaw instance has

• Automated reminders will be sent as you, using your OpenClaw identity
• The cron jobs require OpenClaw to be running and properly configured
• Messages are sent through your configured channels (Telegram/WhatsApp/etc.) using your bot tokens or API credentials

• Automated message sending on your behalf
• Using your existing messaging channel credentials
• Running scheduled tasks that invoke openclaw gateway call message.send

Daily Reminders (Optional)

On first use, offer to set up daily reminders with an interactive flow:

Interactive Setup Flow (Agent Instructions)

On first trigger (when user asks about turnips for the first time):

• Check if already configured:

test -f memory/turnip-config.json && echo "configured" || echo "not configured"

• If NOT configured, offer setup:

• If user says yes, auto-detect config:
• Channel: Extract from inbound context JSON "channel" field (telegram/whatsapp/discord/signal)
• Target ID: Extract from inbound context JSON "sender_id" field
• OpenClaw path: Run which openclaw or use /usr/local/bin/openclaw
• Skill dir: Use the absolute path to this skill's directory
• If auto-detection fails:
• Ask user for missing values explicitly
• Validate before proceeding

{
  "channel": "telegram",
  "target": "8577655544",
  "openclaw_bin": "/usr/local/bin/openclaw",
  "skill_dir": "/home/user/.openclaw/workspace/skills/turnip-prophet",
  "configured_at": "2026-02-23T10:30:00Z"
}

Generate and show cron entries: Show the user exactly what will be added, with their specific values. Example:

# Turnip Prophet reminders for telegram:8577655544
0 8 * * 0 /usr/local/bin/openclaw gateway call message.send --params '{"channel":"telegram","target":"8577655544","message":"🔔 Sunday! Check Daisy Mae'\''s turnip price (90-110 bells) and buy your turnips 🥬"}' 2>&1 | logger -t turnip-prophet
0 12 * * 1-6 /usr/local/bin/openclaw gateway call message.send --params '{"channel":"telegram","target":"8577655544","message":"🔔 Time to check Nook'\''s Cranny turnip prices!"}' 2>&1 | logger -t turnip-prophet
0 20 * * 1-6 /usr/local/bin/openclaw gateway call message.send --params '{"channel":"telegram","target":"8577655544","message":"🔔 Evening price check: Check Nook'\''s Cranny!"}' 2>&1 | logger -t turnip-prophet
45 21 * * 6 /usr/local/bin/openclaw gateway call message.send --params '{"channel":"telegram","target":"8577655544","message":"⏰ FINAL CALL: Turnips expire at 10 PM! Sell now or they'\''ll rot 🗑️"}' 2>&1 | logger -t turnip-prophet

Replace channel/target with detected values. Escape single quotes properly.

Show user what will be stored (before confirmation):

This will save:
• Channel: telegram
• User ID: 8577655544
• Location: memory/turnip-config.json (local only)

This data is stored locally on your machine and is needed so cron reminders can send messages to you.
You can delete the config file anytime with: rm memory/turnip-config.json

Continue? Reply 'confirm' to proceed, or 'cancel' to skip.

On confirmation:

• Write config to memory/turnip-config.json
• Generate a safe install command and show it to the user:

# Save the cron entries to a temp file
   cat > /tmp/turnip-cron-$$.txt <<'TURNIP_EOF'
   [generated entries]
   TURNIP_EOF
   
   # Review the file
   cat /tmp/turnip-cron-$$.txt
   
   # If it looks good, install:
   (crontab -l 2>/dev/null; cat /tmp/turnip-cron-$$.txt) | crontab -

• Ask user to run the commands and confirm when done
• Reply: "✅ Reminders configured. You'll only get pings for missing data. Check crontab -l to verify installation. To remove: rm memory/turnip-config.json and remove the cron entries."

• Reply: "No problem. No data was stored. You can set this up anytime by asking about turnip prices again."

Cron Handler (scripts/cron_handler.sh)

The cron script reads from memory/turnip-config.json for channel/target:

CONFIG_FILE="$SKILL_DIR/memory/turnip-config.json"
if [[ ! -f "$CONFIG_FILE" ]]; then
    echo "Config not found: $CONFIG_FILE" >&2
    exit 1
fi

CHANNEL=$(jq -r '.channel' "$CONFIG_FILE")
TARGET=$(jq -r '.target' "$CONFIG_FILE")
OPENCLAW_BIN=$(jq -r '.openclaw_bin' "$CONFIG_FILE")

Event types:

sunday-daisy:

• Check if memory/turnip-week.json has a buy_price for current week
• If missing: Send reminder "🔔 Sunday! Check Daisy Mae's turnip price (90-110 bells) and buy your turnips 🥬"
• If already set: Stay silent

• Determine which price slot is currently active (Mon AM/PM through Sat PM)
• Check if that slot is filled in memory/turnip-week.json
• If missing: Send reminder "🔔 [Day] [AM/PM]: Check Nook's Cranny turnip prices!"
• If already filled: Stay silent

• Check how many prices are still null
• If any missing OR no buy_price: Send "⏰ FINAL CALL: Turnips expire at 10 PM! Sell now or they'll rot 🗑️"
• If all prices known: "⏰ Last chance to sell turnips tonight!"

How It Works

The skill uses a Python implementation of the actual ACNH turnip price algorithm to predict future prices based on:

• Your Sunday buy price (90-110 bells)
• Any known sell prices from this week
• Previous week's pattern (if known)

• Pattern 0 (Fluctuating): High-low-high-low-high waves
• Pattern 1 (Large Spike): Decreasing then massive spike (up to 6x)
• Pattern 2 (Decreasing): Consistently declining (bad week)
• Pattern 3 (Small Spike): Decreasing then moderate spike (up to 2x)

Usage Instructions

When triggered:

• Read memory/turnip-week.json — get all known data
• Update the file if the user provided a new price
• Run the prediction with all known data
• Generate a chart and send it with the prediction summary

Running the Prediction

echo '{"buy_price": 96, "prices": [84, 81, 78, null, null, null, null, null, null, null, null, null], "previous_pattern": 1}' | python3 scripts/turnip_predict.py

• prices array: [Mon AM, Mon PM, Tue AM, Tue PM, Wed AM, Wed PM, Thu AM, Thu PM, Fri AM, Fri PM, Sat AM, Sat PM]
• Use null for unknown prices

Generating the Chart

After running the prediction, generate a chart image:

python3 scripts/generate_chart.py <buy_price> '<known_json>' '<mins_json>' '<maxs_json>' /tmp/turnip-chart.png

• buy_price: Sunday buy price (integer)
• known_json: array of 12 values, null for unknown (from memory/turnip-week.json prices)
• mins_json: array of 12 min values from the prediction output
• maxs_json: array of 12 max values from the prediction output
• All script paths are relative to the skill directory: skills/turnip-prophet/

Always include the chart with every prediction update.

Presenting Results

Send the chart image via message tool, then reply with a conversational analysis. Don't be robotic — have a personality about it.

Format:

• Chart image with brief caption (buy price, known prices)
• Text reply with:
• Pattern odds as a bullet list with emoji reactions (😬🤞🚀💀 etc.)
• Brief colour commentary — what the data actually means in plain English
• "My take:" — a specific, opinionated recommendation for what to do next (which price to check, when to sell, when to hold)

Pattern odds:
📉 Decreasing: 84.7% 😬
📈 Large Spike: 15.1% 🤞
📊 Small Spike: 0.1%

Not great. Three consecutive drops is strongly pointing to a decreasing week. But there's still a 15% chance of a large spike hiding — if it happens, it'd be Wed-Fri with prices up to 576 bells.

My take: Check the Tuesday PM price. If it drops again, this week is almost certainly a bust — sell and cut your losses. If it jumps up, the spike is on. 🎰

Be direct, be opinionated, skip patterns with 0% probability.

Pattern Descriptions for Users

• Fluctuating (0): Prices go up and down in waves — sell when above 120-130
• Large Spike (1): Prices drop then SPIKE huge (400-600 bells) — wait for it!
• Decreasing (2): Prices keep falling all week — sell ASAP to cut losses
• Small Spike (3): Prices drop then small bump (150-200) — sell during the bump

Error Handling

If the script fails or returns an error:

• Explain what went wrong in simple terms
• Ask user to double-check their input data
• Suggest they try again with corrected information

Notes

• Predictions are probabilistic, not guaranteed
• The algorithm matches the actual game code
• More known prices = better predictions
• Sunday buy price is required
• Previous week's pattern helps but isn't required