AI购物目录修复包

Use this skill after the catalog audit. It converts failed checks into a concise field-level repair queue that a merchant can review and apply in the correct admin or feed tool.

Input and output

• - Input: ai_catalog_audit.json from the companion audit skill.
• Output: ai_catalog_fix_packet.json with priority, product reference, field, reason, evidence, and a human confirmation gate.

Procedure

1. 1. Run the audit first and inspect its score and blockers.
2. Run python scripts/build_catalog_fix_pack.py --audit ai_catalog_audit.json --output ai_catalog_fix_packet.json.
3. A human verifies the source product facts, policies, prices, availability, and channel eligibility before editing or submitting anything.

Boundaries and recovery

The skill does not call Shopify, Google Merchant Center, Catalog API, an AI channel, or any store admin. It never invents missing product facts. Unknown values become HUMAN_FACT_REVIEW. If the audit is malformed, the script exits non-zero; fix the source JSON and rerun.

Acceptance

Every non-passing audit check must create a traceable fix task, while passing products must produce a NO_BLOCKER packet. The packet must show store_write_executed=false.