消费者负反馈诊断

Use this skill when a shop has exported redacted negative feedback from reviews, after-sales, complaints, or customer service and needs a repeatable triage queue.

Prerequisites

• - Python 3.9 or newer.
• A redacted JSON array with feedback references and owner-supplied signal tags.
• A platform profile with the desired response deadline.

Input and output

Each record contains feedbackref, channel, orderref, rating, signaltags, productclaimstatus, refundrequest, and owner_note. Signal tags must be explicit observations; the script does not infer intent from a person\'s identity.

Output is consumerfeedbacktriage.json with cause, severity, evidence tasks, deadline, owner queue, and HUMANREVIEWREQUIRED status.

Procedure

1. 1. Export or enter only redacted feedback facts.
2. Run python scripts/triageconsumerfeedback.py --input examples/feedback.json --output consumerfeedbacktriage.json.
3. Pass the triage output to feedback-reply-guard for a draft queue, or route it to the responsible human team.

Error recovery

Malformed JSON, missing redaction fields, or unsupported rating values exits non-zero. Unknown signal tags remain REVIEW and are never silently treated as a product fault or customer fault.

Permission and privacy boundaries

The skill does not contact a consumer, post a reply, decide a refund, admit liability, reveal private data, or change an order. Keep order references pseudonymous and remove names, phone numbers, addresses, account IDs, and chat screenshots before input.

Acceptance

The example must create separate queues for a product-claim issue, a logistics issue, and an unknown issue, with evidence tasks and HUMANREVIEWREQUIRED on every row.