电商宣传风险审计

Use this skill before a product title, description, image text, live script, short-video script, or comment is handed to a human publisher.

Prerequisites

• - Python 3.9 or newer.
• A redacted claim manifest with the content channel, evidence references, owner fact status, and risk signals.
• Current platform rules and the owner\'s approved product facts remain the source of truth.

Input and output

Each row contains claimref, channel, text, evidencerefs, productfactstatus, priceorguaranteestatus, brandvisibility, offplatformsignal, piisignal, and ownernote.

Output is claimriskaudit.json with risk flags, severity, evidence tasks, and fixed externaleditexecuted=false.

Procedure

1. 1. Copy only redacted claim facts into JSON.
2. Run python scripts/auditclaimrisk.py --input examples/claims.json --output claimriskaudit.json.
3. Give the audit to claim-remediation-kit and a human compliance or operations owner.

Error recovery

Malformed JSON or missing risk fields exits non-zero. Unknown facts become REVIEW. The script never treats a keyword hit as a final legal or platform decision.

Permission and privacy boundaries

This skill does not edit copy, remove watermarks, publish content, contact users, access accounts, or certify legal compliance. Remove names, phone numbers, addresses, account identifiers, and private customer text before input.

Acceptance

The example must flag an unsupported absolute claim, a brand-masking signal, and an off-platform contact signal, with evidence tasks and human review on every item.