L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Productivité · SKILL.md

manage-tcg-collection

manage-tcg-collection

Organize, track, and value a trading card game collection. Covers inventory methods, storage best practices, grade-based valuation, want-list management, and collection analytics for Pokemon, MTG, Flesh and Blood, and Kayou cards. Use when starting a new collection and setting up inventory tracking, cataloging an existing collection that has grown beyond ca…

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: manage-tcg-collection locale: caveman-ultra source_locale: en source_commit: 82c77053 fence_basis_commit: 82c77053 translator: "Julius Brussee homage — caveman" translation_date: "2026-04-26" description: > Organize, track, and value a trading card game collection. Covers inventory methods, storage best practices, grade-based valuation, want-list management, and collection analytics for Pokemon, MTG, Flesh and Blood, and Kayou cards. Use when starting a new collection and setting up inventory tracking, cataloging an existing collection that has grown beyond casual knowledge, valuing a collection for insurance or sale, or deciding which cards to submit for professional grading based on value potential. license: MIT allowed-tools: Read Grep Glob WebFetch WebSearch metadata: author: Philipp Thoss version: "1.0" domain: tcg complexity: basic language: natural tags: tcg, collection, inventory, storage, valuation, pokemon, mtg, fab, kayou

Manage TCG Collection

Organize, inventory, value TCG collection → structured tracking + storage + data-driven valuation.

Use When

  • New collection → set up tracking
  • Existing collection grew past casual knowledge
  • Value for insurance/sale/estate
  • Manage want-lists + trade binders
  • Decide grading candidates by value potential

In

  • Required: Game(s) (Pokemon, MTG, FaB, Kayou)
  • Required: Scope (entire / sets / specific cards)
  • Optional: Current system (sheet, app, binder)
  • Optional: Goal (sets, play, invest, nostalgia)
  • Optional: Budget for supplies/grading

Do

Step 1: Inventory System

Tracking system matched to size.

  1. Method by size:
Collection Size Guide:
+-----------+-------+-------------------------------------------+
| Size      | Cards | Recommended System                        |
+-----------+-------+-------------------------------------------+
| Small     | <200  | Spreadsheet (Google Sheets, Excel)         |
| Medium    | 200-  | Dedicated app (TCGPlayer, Moxfield,        |
|           | 2000  | PokeCollector, Collectr)                   |
| Large     | 2000+ | Database + app combo with barcode scanning |
+-----------+-------+-------------------------------------------+
  1. Fields per card:
    • Identity: Set, num, name, variant (holo, reverse, full art)
    • Condition: Raw grade (NM, LP, MP, HP, DMG) or numeric
    • Qty: Copies owned
    • Location: Binder page, box label, slab
    • Acquisition: Date, price paid, source
    • Value: Market value at condition, last update
  2. Set up system w/ fields
  3. Cadence (weekly active, monthly stable)

→ Working system + fields, ready for entry. Matches scale.

If err: ideal app missing → sheet. Consistency > format. Sheet updated > app abandoned.

Step 2: Catalog

Enter cards into system.

  1. Sort physically first:
    • By set
    • Within set, by num ascending
    • Variants w/ base
  2. Enter:
    • Bulk where possible (barcode, set checklists)
    • Honest condition — over-grade → valuation err
    • Note provenance (signed, 1st ed, prizes)
  3. Large → sessions:
    • One set/box per session
    • Mark progress
    • Random sample verify
  4. Cross-ref set checklists → completion %

→ All cards entered w/ condition + location. Completion % known.

If err: too large → priority: rares first, bulk-enter commons w/ est qty. 80% accurate > none.

Step 3: Physical Storage

Store by value + use.

  1. Storage tier system:
Storage Tiers:
+----------+---------------+----------------------------------------------+
| Tier     | Card Value    | Storage Method                               |
+----------+---------------+----------------------------------------------+
| Premium  | >$50          | Top-loader + team bag, or penny sleeve in    |
|          |               | magnetic case. Stored upright in a box.       |
| Standard | $5-$50        | Penny sleeve + top-loader or binder with      |
|          |               | side-loading pages.                          |
| Bulk     | <$5           | Row box (BCW 800-count or similar), sorted    |
|          |               | by set. No individual sleeves needed.         |
| Graded   | Any (slabbed) | Upright in graded card box. Never stack heavy.|
+----------+---------------+----------------------------------------------+
  1. Env controls:
    • Cool, dry, dark (not attic/basement)
    • No direct sun, humidity, temp swings
    • Silica gel in boxes
  2. Label all:
    • Box: contents (set, range, date)
    • Binder pages → location codes
    • Graded: ID matches digital
  3. Update locations in system

→ All cards stored by tier w/ location in inventory. Premium protected, bulk accessible.

If err: no premium supplies → penny sleeve + top-loader min for >$10. Upgrade as available.

Step 4: Value Collection

Current market values.

  1. Pricing source:
    • TCGPlayer Market Price: US (MTG, Pokemon)
    • CardMarket: EU
    • eBay Sold Listings: Rare/unique
    • PSA/BGS Price Guide: Graded
  2. Update Standard + Premium values
  3. Bulk → per-set bulk pricing
  4. Summary:
Collection Value Summary:
+------------------+--------+--------+
| Category         | Count  | Value  |
+------------------+--------+--------+
| Graded cards     |        | $      |
| Premium ungraded |        | $      |
| Standard cards   |        | $      |
| Bulk cards       |        | $      |
+------------------+--------+--------+
| TOTAL            |        | $      |
+------------------+--------+--------+
  1. Grading candidates: grade-premium > grading cost
    • Rule: grade if (graded value - raw) > 2x grading cost

→ Current valuation, per-card for significant + aggregate bulk. Grading candidates flagged.

If err: stale pricing → note date + source. Rare → multi-source median. Never trust outlier.

Step 5: Maintain + Optimize

Ongoing routines.

  1. Updates (Step 1 cadence):
    • New acquisitions immediately
    • Premium quarterly, Standard semi-annually
    • Re-tier as values shift
  2. Want-list:
    • Desired cards + max price
    • Cross-ref vs. trade binder
    • Price alerts if supported
  3. Analytics:
    • Monthly value snapshots
    • Set completion %
    • Concentration risk (one card/set heavy)
  4. Annual audit:
    • Physical vs. inventory random sample
    • Storage check (humidity, pests)
    • Re-eval grading candidates

→ Living system, current, supports buy/sell/grade/trade decisions.

If err: lapse → Premium values first, then new acquisitions. Know top values today.

Check

  • System established w/ fields
  • All cards cataloged + condition + location
  • Storage matches tiers
  • Env controls (cool, dry, dark)
  • Valued w/ current prices + dates
  • Grading candidates flagged w/ cost/benefit
  • Maintenance cadence followed
  • Want-list maintained

Traps

  • Over-grade own cards: Collectors rate own +1-2 grades. Honest or use grade-tcg-card
  • Ignore bulk: 800 commons @ $0.10 = $80. Track
  • Bad env: Humidity + temp swings damage > handling. Env > sleeves
  • Stale values: 6 months old → wildly off, esp. set release/ban
  • No backup: Export CSV monthly. Photograph premium for insurance

→

  • grade-tcg-card — structured grading
  • build-tcg-deck — deck from inventory
Skills enregistrées