[Community Submission] Add skill: variant-option-normalizer

[raqibdev]

Description

Detects inconsistent Shopify variant option data and produces a corrected CSV plus a change log. Covers 8 detection categories across a 3-turn merchant conversation: audit, confirm plan,
produce output.

What's in this PR:

• Core skill (SKILL.md) with full detection logic, conversation flow, edge cases, and error handling
• Python audit script (normalize_audit.py) for deterministic checks (whitespace, case, duplicates, missing images, size ordering, incomplete variant matrix detection)
• Alias asset files (size_aliases.json, option_name_aliases.json) with expanded synonym maps and size system classifications
• JSON output schema reference (references/json-schema.md) documenting every field the script emits
• Example output (references/example-output.md) showing a full 3-turn run with audit report, normalization plan, corrected CSV, and change log
• Fixture: fixtures/greatoutdoorsco/shopify-products-messy-variants.csv covering all 8 issue types across 5 products

Key decisions made during development:

• Confidence framing applied to all 8 detection categories, not just aliases, to guide how findings are presented to the merchant
• Post-normalization collision check added to Turn 2 to catch cases where two distinct values would merge into the same canonical form
• Fill-by-color image proposal added to Turn 2, sourcing sibling variant URLs rather than leaving gaps
• approved_mapping.json export offered at Turn 3 so future runs on the same store can skip re-analysis for already-resolved values
• LLM semantic checks scoped to within the same option_name group only, to prevent cross-slot false positives
• Error handling section added for malformed CSV, missing required columns, wrong delimiters, and unrecoverable ambiguity
• Detection Categories section trimmed and pointed to json-schema.md to avoid duplication

Contributor Attestation

• I agree to the Terms of Service and Privacy Policy, and I understand that if my submission is accepted it will be published under the Apache 2.0 license. I confirm I have the rights to submit this content and it does not include confidential information or personal data.