Turn the refunds email thread into a counter procedure
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.
Dorothy keeps mixing up how refunds work. Please turn the email thread into one written procedure we can print for the register and save it as procedure.md.
Workspace
- email_thread_refunds.txt
- return_policy_posted.pdf
Expected deliverables
- procedure.md
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 |
|---|---|---|---|
| approval threshold | text_sentence_matches | yes | 0 / 6 |
| refund window | text_sentence_matches | yes | 0 / 6 |
| store credit window | text_sentence_matches | yes | 0 / 6 |
| cash refunds | text_sentence_matches | yes | 0 / 6 |
| refunds logged in the sheet | text_sentence_matches | yes | 0 / 6 |
| steps in the right order | custom | yes | 2 / 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.
- approval threshold —
text_sentence_matches, required - refund window —
text_sentence_matches, required - store credit window —
text_sentence_matches, required - cash refunds —
text_sentence_matches, required - refunds logged in the sheet —
text_sentence_matches, required - steps in the right order —
custom, required
Recorded results, desk release
This is one of the 7 tasks with a task-specific equivalence grader in the frozen scorer, intended to accept equivalent representations. Whether an equivalence rule matches the business requirement remains open to review. Original and frozen verdicts can differ; see what the frozen scorer changed.
| System | Repetitions | Frozen / 3 | Original / 3 | Median minutes | Mean model estimate (USD) |
|---|---|---|---|---|---|
| Proto / DeepSeek V4.1 Flash | 2 | 1 | 0.6 | 0.0090 | |
| Codex / GPT-5.6 sol | 2 | 2 | 1.3 | 0.1958 |
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": "sop-from-thread".
Recorded results, desk comparison, 28 September 2026
| System | Repetitions | Frozen / 3 | Original / 3 | Median minutes |
|---|---|---|---|---|
| Proto / GPT-6 sol | 0 | 0 | 1.2 | |
| Codex / GPT-6 sol | 1 | 0 | 2.8 | |
| Proto / DeepSeek V4.1 Flash | 2 | 2 | 0.4 |
A separate campaign under the same frozen scorer, not pooled with the release. Conditions and data.