queue-fairness-log

---
name: queue-fairness-log
description: Builds the record that proves a waitlist was run in arrival order rather than by who the host liked the look of, while keeping the personal data in it to only what the UK GDPR allows. Logs name, party size, join time and seated time against the order actually seated, flags any seating that jumped the queue with no recorded reason, and sets a retention date so names don't sit in a notebook forever. Use when you want to prove a waitlist was run fairly, and when deciding what to keep about queuing guests and for how long.
---

# A waitlist record that proves who went first

Give it the people who joined the queue, in order, with party size and seated time. Get back a log showing the order they should have been seated in against the order they actually were, any jump flagged, and a retention date for deleting the names.

## What it does

1. **Log only what running the queue needs.** UK GDPR data minimisation requires data to be "adequate, relevant and limited to what is necessary in relation to the purposes for which they are processed." A name or contact method, party size, and two timestamps is all a fair waitlist needs.

2. **Record join time separately from seated time.** The gap between the two is the entire evidence of fairness, and it only exists if both are kept.

3. **Flag a jump, and require a reason.** A party seated earlier than their join order predicts is a jump; a one-line reason is required. A jump with no reason is the one thing this skill treats as a finding.

4. **Set a retention date before collecting, not after a complaint.** Storage limitation requires data "kept... for no longer than is necessary for the purposes for which the personal data are processed." A waitlist name has done its job once the shift ends.

5. **Keep the aggregate, drop the names.** After the retention date, keep counts (queued, average wait, jump count); the names are not needed for any of that.

6. **Write the log so a stranger could see the order was fair**, or see exactly where it was not.

## Then it checks

1. Every row has only a name/contact, party size, join time and seated time.
2. Join order and seated order are both present and comparable.
3. Every jump has a one-line reason recorded.
4. A retention date is set, and no name sits past it.
5. Aggregate counts can be produced without the names.

Any check fails: redo that step once. Failed twice: say what is wrong and stop.

## Rules
- Public information only. Never invent a fact, number or quote. Anything sent in someone's name says whose name it is.
- Never log more than a name/contact and party size; no notes on appearance or behaviour.
- Never let a name sit past its retention date, or record a jump with no reason.
- Working document for the owner's data protection adviser to confirm retention and lawful basis.

## Built from
- UK GDPR, Article 5(1)(c), legislation.gov.uk/eur/2016/679/article/5, read 4 Nov 2026: step 1.
- UK GDPR, Article 5(1)(e), legislation.gov.uk/eur/2016/679/article/5, read 4 Nov 2026: steps 4 and 5.

# queue-fairness-log

## You are given
A list of queue or waitlist entries for a service period, each with: a name or contact method, party size, join time, and seated time (or `not seated` if the party left or was never called). The retention period in days, already decided by Claude or the owner. Any reasons already recorded for a party being seated out of join order.

## Produce
Write into a `./queue-fairness-log-output/` folder:

1. `queue-log.csv` with columns: `row_ref`, `name_or_contact`, `party_size`, `join_time`, `seated_time`, `outcome`. `row_ref` is Q0001 upward. `outcome` is exactly one of `seated`, `left before seated`, `never called`. No other field about the guest is recorded in this file.
2. `order-check.csv` with columns: `row_ref`, `join_order_position`, `seated_order_position`, `positions_jumped`, `reason_recorded`, `reason_text`. `join_order_position` and `seated_order_position` are both ranked 1 upward by time within the same service period. `positions_jumped` is `join_order_position` minus `seated_order_position`; a positive number means the party was seated earlier than their join order would predict. `reason_recorded` is `yes` or `no`.
3. `retention.csv` with columns: `row_ref`, `join_time`, `retention_period_days`, `delete_on_date`, `name_still_present`. `delete_on_date` is `join_time` plus `retention_period_days`. `name_still_present` is `yes` or `no`, checked against today's date.
4. `aggregate-summary.csv` with columns: `service_period`, `total_queued`, `total_seated`, `total_left_before_seated`, `average_wait_minutes`, `jump_count`, `jump_count_with_reason`, `jump_count_without_reason`. This file contains no names and can be kept after the retention date.
5. `gaps.md` - a numbered list of: any row with `positions_jumped` greater than zero and `reason_recorded` of `no`; any row past its `delete_on_date` with `name_still_present` of `yes`; any row logging more than name or contact, party size, and the two timestamps; any row with a missing join or seated time where the outcome is `seated`.

## Rules
- Codex counts, ranks and records. It never decides the retention period, never writes a reason for a jump, and never adds a field about a guest beyond what it was given.
- Positions and jumps are calculated exactly from the supplied timestamps, never estimated.
- Never carry a name into `aggregate-summary.csv`.
- Use British English and DD Month YYYY dates. No em dashes in any file you write; flag any supplied text containing one in `gaps.md`.
- Every file ends with this line: this is a working document prepared for the owner to check against their own data protection policy, and for their data protection adviser to confirm retention periods and lawful basis. It is not legal advice.

## Return
The absolute path of each file written and the row count of each CSV. The total queued, seated, and left-before-seated counts. The jump count, with and without a recorded reason, and the row references of any jump with no reason. The number of rows past their retention date with a name still present. The `gaps.md` item count.