# Support queue handoff

## Deliverables
- `triage.csv`: decisions for all 10 tickets, using the requested seven columns.
- `replies.md`: one empathetic draft per ticket, each under 120 words including its source note.
- `checks.md`: actual local validation results and word counts.
- `build_and_check.py`: standard-library draft generator and baseline checks.
- `revise_submission.py`: reproduces the revision from the archived original and runs expanded on-disk validation.
- `review_checks.md`: expanded checks and evidence review.
- `CHANGELOG.md`: brief revision summary.
- `original_submission/`: byte-preserved original deliverables and SHA-256 manifest.

## Scope and conventions
Assessment date: **October 20, 2026** (`PROMPT.md`, Task 7). Policy authority: `inputs/policy.md`, Alder Supply policy — version October 1, 2026. All tickets are assessed as requests/reports on October 20 because separate ticket timestamps are absent; T701 explicitly specifies that date (`inputs/tickets.csv`, T701). Calendar-day elapsed values are assessment date minus delivery/purchase date; day 30 is included for standard returns (`inputs/policy.md`, Returns).

`refund_cad` is an assessment amount, **not an issued or unconditionally approved payment**:
- Positive amounts on conditional rows are contingent merchandise/duplicate amounts, with prerequisites in `reason`.
- T703's CAD 60 is the alternative to the requested replacement, not an additional refund.
- `0.00` means no refund under the applicable policy on current evidence.
- Blank means undetermined: T708 is safety-first with no compensation decision; T710 has no matching order.
- Taxes and shipping cannot be calculated from these inputs (`inputs/policy.md`, Evidence). Do not total these rows as approved refunds.

Priority labels are workflow assumptions, not additional store policy: **P0** highest, immediate human safety review; **P1** next, payment-team duplicate verification; **P2** routine. Only the highest-priority safety requirement is prescribed by policy (`inputs/policy.md`, Safety).

`needs_human=yes` means substantive receipt/photo/payment verification, order investigation or safety review remains required. `no` means the supplied evidence supports the policy decision without special investigation; it does not authorize sending or processing the draft automatically. T701 still needs receipt of unopened goods, T703 needs photo verification, T706 needs payment-team verification, T708 needs urgent safety review, and T710 needs order identification (`inputs/policy.md`, Returns; Defects and duplicates; Safety; Evidence; respective ticket/order IDs in triage).

## Outstanding work and limitations
- **T708: urgent human safety escalation is required but has not been executed in this drafting exercise.** Keep the adapter unplugged, prohibit retesting and request order confirmation/contact privately. Do not delay escalation for evidence or a refund decision (`inputs/policy.md`, Safety; `inputs/tickets.csv`, T708; `inputs/orders.csv`, A708).
- Receive and verify T701's unopened return before refund; obtain and verify T703's damage photo; verify T706's duplicate with the payment team. No receipt, inspection, reversal, prepaid label or shipment is evidenced (`inputs/policy.md`, Returns; Defects and duplicates; Evidence).
- T710 cannot be matched to an order from `UNKNOWN`. Do not infer it is another customer's headphones order or disclose others' details (`inputs/tickets.csv`, T710; `inputs/orders.csv`, no UNKNOWN match; `inputs/policy.md`, Safety; Evidence).
- T704's manager-approval claim has no supporting record and is not accepted as an override (`inputs/tickets.csv`, T704; `inputs/orders.csv`, A704; `inputs/policy.md`, Returns; Evidence).
- This is draft-only work. No replies sent, refunds issued, escalations filed or customer/payment contacts made in preparing this submission. This does not establish historical action status outside the supplied records. Supplied files are the only evidence.

## Checks and estimated human effort
The executed Python checks cover ticket completeness, CSV schema, reply lengths, calendar arithmetic, course/payment facts, citations and order-ID isolation. See `checks.md` and `review_checks.md`; the ID check alone is not a substitute for human privacy review.

Estimated human effort: **15–25 active minutes** to review these drafts, initiate the safety escalation and request outstanding evidence/payment verification. This is an estimate, not measured time. Final resolution also depends on customer responses, return receipt and payment/safety review; elapsed completion time cannot be estimated from the inputs.
