Tie out the Keystone supply statement to our books
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.
Keystone's August statement is in the folder with our ledger for them. Reconcile the two before we pay, Luis's note has how he likes it done. Save it as vendor_recon.csv.
Workspace
- ap_vendor_ledger_keystone.csv
- keystone_statement_2026-08.pdf
- note_statement_recon.txt
Expected deliverables
- vendor_recon.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 |
|---|---|---|---|
| note's columns | csv_columns | yes | 0 / 6 |
| one row per open item | csv_set_equal | yes | 0 / 6 |
| row count | csv_row_count | yes | 0 / 6 |
| status per document | csv_values_match | yes | 0 / 6 |
| difference per document | csv_values_match | yes | 0 / 6 |
| variance ties to the balances | custom | yes | 0 / 6 |
Current task declaration (6 checks)
From task.yaml. Original runner grading and frozen retrospective scoring are separate. Read the frozen source for the rules behind the reported results.
- note's columns —
csv_columns, required - one row per open item —
csv_set_equal, required - row count —
csv_row_count, required - status per document —
csv_values_match, required - difference per document —
csv_values_match, required - variance ties to the balances —
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 | 1.8 | 0.0244 | |
| Codex / GPT-5.6 sol | 3 | 3 | 1.3 | 0.1895 |
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": "vendor-statement-reconcile".
Recorded results, desk comparison, 28 September 2026
| System | Repetitions | Frozen / 3 | Original / 3 | Median minutes |
|---|---|---|---|---|
| Proto / GPT-6 sol | 3 | 3 | 1.5 | |
| Codex / GPT-6 sol | 3 | 3 | 1.2 | |
| Proto / DeepSeek V4.1 Flash | 3 | 3 | 0.5 |
A separate campaign under the same frozen scorer, not pooled with the release. Conditions and data.