Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Look up and refund Google Play orders by order ID with gplay `orders` — admin diagnostics, not runtime purchase-token verification. Use when a buyer complaint or payout report hands you an order ID and you need its state, or when issuing a money-moving `--confirm`-gated refund.
.claude/skills/pollyglot-gplay-orders/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 42 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-18 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -66% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -47% | 0% |
gplay orders is the admin commerce surface: look up an order by its order ID, and, when warranted, refund it. Both commands are [experimental]. Shared conventions are in gplay-cli-usage.
orders viewbashgplay orders view GPA.1234-5678-9012-34567 # orders.get gplay orders view GPA.1111-... GPA.2222-... # orders.batchget: one package per call, all-or-nothing gplay orders view GPA.1234-... --output json # full Order, verbatim, with the fields the summary omits
orders refund (money-moving, irreversible)bashgplay orders refund GPA.1234-5678-9012-34567 --dry-run # rehearse, no HTTP gplay orders refund GPA.1234-5678-9012-34567 --confirm # money back, access kept gplay orders refund GPA.1234-5678-9012-34567 --revoke --confirm # money back, entitlement revoked
Pick --revoke before running: it is the only choice the command leaves to you (money back and access kept, or money back and entitlement revoked, a subscription's future billing stopping too). One order per call.
Money capabilities live in no Role bundle (viewer … admin): grant them explicitly (gplay-team). orders view needs CAN_VIEW_FINANCIAL_DATA, orders refund needs CAN_MANAGE_ORDERS; a 403 (exit 11) names the missing one.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | pass→pass | 22,634 | 12,742 | -44% | 1 | 1 | 0% | 2,560 | 788 | -69% | 0 | 0 | — |
case-18 | fail→pass | 18,070 | 7,341 | -59% | 1 | 1 | 0% | 2,039 | 749 | -63% | 0 | 0 | — |
case-01 | fail→pass | 10,752 | 9,041 | -16% | 1 | 1 | 0% | 1,525 | 1,023 | -33% | 0 | 0 | — |
case-02 | fail→pass | 28,662 | 14,048 | -51% | 1 | 1 | 0% | 1,802 | 1,507 | -16% | 0 | 0 | — |
case-03 | fail→pass | 42,349 | 12,538 | -70% | 1 | 1 | 0% | 2,884 | 968 | -66% | 0 | 0 | — |
case-04 | fail→pass | 13,618 | 6,748 | -50% | 1 | 1 | 0% | 2,192 | 1,171 | -47% | 0 | 0 | — |
case-05 | fail→pass | 29,638 | 5,789 | -80% | 1 | 1 | 0% | 2,150 | 649 | -70% | 0 | 0 | — |
case-06 | pass→pass | 20,765 | 17,251 | -17% | 1 | 1 | 0% | 2,367 | 2,405 | +2% | 0 | 0 | — |
case-07 | pass→pass | 28,091 | 9,868 | -65% | 1 | 1 | 0% | 1,778 | 1,065 | -40% | 0 | 0 | — |
case-08 | fail→pass | 51,364 | 2,971 | -94% | 1 | 1 | 0% | 2,567 | 812 | -68% | 0 | 0 | — |
case-09 | fail→pass | 28,355 | 8,346 | -71% | 1 | 1 | 0% | 1,695 | 956 | -44% | 0 | 0 | — |
case-10 | fail→pass | 8,130 | 10,429 | +28% | 1 | 1 | 0% | 1,196 | 1,376 | +15% | 0 | 0 | — |
case-11 | fail→pass | 19,453 | 8,455 | -57% | 1 | 1 | 0% | 2,200 | 769 | -65% | 0 | 0 | — |
case-12 | fail→pass | 19,235 | 8,684 | -55% | 1 | 1 | 0% | 2,762 | 750 | -73% | 0 | 0 | — |
case-13 | fail→pass | 16,816 | 6,489 | -61% | 1 | 1 | 0% | 1,730 | 665 | -62% | 0 | 0 | — |
case-14 | pass→pass | 26,397 | 9,226 | -65% | 1 | 1 | 0% | 1,778 | 1,034 | -42% | 0 | 0 | — |
case-15 | fail→pass | 53,821 | 7,663 | -86% | 1 | 1 | 0% | 2,867 | 772 | -73% | 0 | 0 | — |
case-16 | fail→pass | 18,207 | 5,950 | -67% | 1 | 1 | 0% | 2,168 | 756 | -65% | 0 | 0 | — |
case-19 | pass→pass | 8,449 | 10,373 | +23% | 1 | 1 | 0% | 1,179 | 856 | -27% | 0 | 0 | — |
case-20 | fail→pass | 11,741 | 2,698 | -77% | 1 | 1 | 0% | 1,671 | 790 | -53% | 0 | 0 | — |
case-21 | fail→pass | 18,338 | 15,212 | -17% | 1 | 1 | 0% | 1,787 | 916 | -49% | 0 | 0 | — |
case-22 | pass→pass | 14,132 | 8,033 | -43% | 1 | 1 | 0% | 2,183 | 724 | -67% | 0 | 0 | — |
case-23 | pass→pass | 15,348 | 4,720 | -69% | 1 | 1 | 0% | 1,582 | 1,074 | -32% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +70 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/23/2026 | +74% |
| gemini-3.6-flash | verified | 8/4/2026 | +68% |
Other measured skills in the registry, with their headline benchmark lift.