Weekly hours audit from time-clock punches
The request
The task request supplied to the adapter, alongside the workspace below. Runtime instructions, authorized tools and skills also affect the system. Containerized runs withhold host-side references; local mode is not an isolation boundary.
Bea needs last week's hours from the time clock before she runs payroll. Turn the punch export into hours_audit.csv; her note says how she counts shifts.
Workspace
- note_from_bea.txt
- schedule_week_of_aug31.xlsx
- timeclock_export_2026-08-31_to_09-07.csv
Expected deliverables
- hours_audit.csv
This task also has a task-specific grader (check.py).
Recorded acceptance checks
The table uses check names and outcomes from the frozen campaign ledger. Acceptance requires all required predicates to pass; a predicate may itself contain a tolerance or coverage threshold. Its scope is part of the measurement contract.
| Recorded check | Type | Required | Failed / observed |
|---|---|---|---|
| requested columns | csv_columns | yes | 0 / 6 |
| row count | csv_row_count | yes | 0 / 6 |
| daily hours per person | custom | yes | 0 / 6 |
| missing punches flagged | custom | yes | 0 / 6 |
Current task declaration (4 checks)
From task.yaml. Original runner grading and frozen retrospective scoring are separate. Read the frozen source for the rules behind the reported results.
- requested columns —
csv_columns, required - row count —
csv_row_count, required - daily hours per person —
custom, required - missing punches flagged —
custom, required
Recorded results, desk release
| System | Repetitions | Frozen / 3 | Original / 3 | Median minutes | Mean model estimate (USD) |
|---|---|---|---|---|---|
| Proto / DeepSeek V4.1 Flash | 3 | 3 | 2.2 | 0.0276 | |
| Codex / GPT-5.6 sol | 3 | 3 | 2.2 | 0.2253 |
These are repeated executions on the released fixture, not independent new business cases. Cost is captured model usage priced under the recorded assumptions, excluding infrastructure and human review. Per-attempt check verdicts, raw and frozen verdicts, usage, receipt and artifact hashes for this task are in results/latest/attempts.jsonl under "task": "clock-punch-audit".
Recorded results, desk comparison, 28 September 2026
| System | Repetitions | Frozen / 3 | Original / 3 | Median minutes |
|---|---|---|---|---|
| Proto / GPT-6 sol | 3 | 3 | 2.1 | |
| Codex / GPT-6 sol | 3 | 3 | 1.2 | |
| Proto / DeepSeek V4.1 Flash | 3 | 3 | 1.5 |
A separate campaign under the same frozen scorer, not pooled with the release. Conditions and data.