Card Evaluation

❌ NEVER:

• Make recommendations based on card names alone
• Assume what a card does without reading oracle text
• Skip reading cache because "it's probably just a simple card"
• Use price as the primary decision factor

Card Validation (MANDATORY BEFORE RECOMMENDING)

# Validate color identity BEFORE presenting recommendations
npm run validate-cards -- <deck-file> "Card 1" "Card 2" "Card 3"

# Example output:
# ❌ Thassa's Oracle - color identity [U] not in commander [B, W]
# ✅ Mana Crypt - colorless, legal
# ✅ Opposition Agent - [B], legal in [B, W]

After validation passes, also:

1. Fetch price: npm run price-cards "Card 1" "Card 2" (batch up to 15)
2. Check collection: npm run collection find "Card Name"
3. Read oracle text if needed: npm run fetch "Card Name" or npm run rulings "Card Name"
4. ONLY THEN suggest

🚨 NEVER HALLUCINATE CARD ABILITIES OR COSTS - Read fetched oracle text and verify CMC from output.

Color Identity Validator

npm run validate-colors <deck-file>    # Requires cache

• Checks every card's color identity against the commander
• Reports all violations with specific colors
• Returns exit code 1 if violations found

Card Counter

npm run count <deck-file>
npx tsx packages/mtg-validation/src/cards/card-counter.ts <deck-file> --validate  # Validate category headers

• Counts total and unique cards
• Breaks down by category
• Validates headers match actual counts
• ALWAYS use before finalizing deck recommendations

Deck Reduction Workflow

When user asks to reduce deck size:

1. Read commander oracle text FIRST - understand transformative effects:

   • Does commander make creatures ALL types? → Tribal lords are GOOD
   • Does commander make lands ALL types? → Land-type-matters cards are GOOD
   • Does commander change fundamental rules?

2. Check EDHREC data SECOND (mandatory for every proposed cut):

   npm run edhrec "Commander Name" -- --synergy
   npm run edhrec "Commander Name" -- --category creatures
   npm run edhrec "Commander Name" -- --category instants
   npm run edhrec "Commander Name" -- --category sorceries
   

   • Top 20 in category → PROTECTED, DO NOT CUT
   • Top 50 → Requires user approval
   • Document EDHREC ranking in cut justification

3. Systematic synergy check:

   • For EACH card, ask: Does it interact with commander's key abilities?
   • Does it say "if you control a [type]"? → Commander may enable it
   • Does it care about counters/types? → Commander may synergize
   • Check the LAST LINE of oracle text - don't miss abilities

4. Prefer cutting expensive unowned cards (check pricing and collection data)

Example mistakes to avoid:

• ❌ Cutting tribal lords when commander makes all creatures all types
• ❌ Cutting "conditional" cards without checking if commander enables them
• ❌ Cutting cards based on names alone without reading oracle text
• ❌ Removing high-synergy EDHREC cards

Card Search & Data Commands

Manual Deck Import (User-Provided Card Lists)

When user provides a raw card list (not a Moxfield URL):

1. ALWAYS fetch card data first: npm run fetch "Card 1" "Card 2" ...
2. Run import validator: npm run validate-import /tmp/cards.txt "Commander Name" commander
3. Review validation output - actual card types, EDHREC synergy, card count
4. NEVER remove high-synergy cards identified by the validator
5. Create deck folder with timestamp, verify 100 cards with npm run count

Common type-guessing mistakes:

• ❌ "Sage of the Maze" sounds like an enchantment → Actually a Creature
• ❌ "Lord of the Unreal" sounds like an enchantment → Actually a Creature
• ✅ Always verify type_line from fetch output