Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Push and validate the app's Google Play Data Safety declaration from a versioned CSV — the only Play compliance surface with an API (content rating and the other declarations stay Console-manual). Use when updating the Data Safety form or gating it in CI.
.claude/skills/pollyglot-gplay-compliance/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -73% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -52% | 0% |
Manage the app's Data Safety declaration from a canonical CSV. Shared conventions are in gplay-cli-usage.
Content rating (IARC), target audience/Families, ads, news-app, government-app and the other declarations stay manual in Play Console: list them as a manual step in the release runbook.
The declaration is write-only (ADR-0014): a direct POST outside the Edits model that replaces the whole document. gplay cannot read it back; there is no get, only set and an offline validate. validate passing means well-formed, not accepted: only the live set POST arbitrates against Google's schema.
bashgplay compliance datasafety validate # offline structural check, no auth gplay compliance datasafety set --dry-run # rehearse: validate + resolve target + size, no HTTP gplay compliance datasafety set --confirm # the real write, replaces the live declaration
bashgplay compliance datasafety validate || exit $? # offline gate, fails on bad CSV gplay compliance datasafety set --confirm --output json # publish from versioned CSV
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,536 | 22,325 | +4% | 1 | 1 | 0% | 3,577 | 1,328 | -63% | 0 | 0 | — |
case-02 | pass→pass | 82,633 | 13,763 | -83% | 1 | 1 | 0% | 3,208 | 2,475 | -23% | 0 | 0 | — |
case-03 | fail→pass | 21,392 | 11,172 | -48% | 1 | 1 | 0% | 2,568 | 915 | -64% | 0 | 0 | — |
case-04 | fail→pass | 31,628 | 9,651 | -69% | 1 | 1 | 0% | 3,152 | 847 | -73% | 0 | 0 | — |
case-05 | pass→pass | 16,727 | 9,882 | -41% | 1 | 1 | 0% | 1,616 | 722 | -55% | 0 | 0 | — |
case-06 | pass→pass | 16,847 | 9,558 | -43% | 1 | 1 | 0% | 2,138 | 947 | -56% | 0 | 0 | — |
case-07 | pass→pass | 20,210 | 9,445 | -53% | 1 | 1 | 0% | 2,185 | 764 | -65% | 0 | 0 | — |
case-08 | fail→pass | 13,679 | 9,029 | -34% | 1 | 1 | 0% | 1,823 | 829 | -55% | 0 | 0 | — |
case-09 | fail→pass | 30,011 | 9,646 | -68% | 1 | 1 | 0% | 1,803 | 869 | -52% | 0 | 0 | — |
case-10 | pass→pass | 14,767 | 9,726 | -34% | 1 | 1 | 0% | 1,328 | 981 | -26% | 0 | 0 | — |
case-11 | pass→pass | 17,512 | 13,277 | -24% | 1 | 1 | 0% | 1,655 | 1,521 | -8% | 0 | 0 | — |
case-12 | fail→fail | 6,806 | 2,269 | -67% | 1 | 1 | 0% | 990 | 571 | -42% | 0 | 0 | — |
case-13 | pass→pass | 14,534 | 3,625 | -75% | 1 | 1 | 0% | 1,298 | 699 | -46% | 0 | 0 | — |
case-14 | fail→pass | 60,843 | 20,165 | -67% | 1 | 1 | 0% | 2,282 | 996 | -56% | 0 | 0 | — |
case-15 | fail→pass | 63,143 | 7,901 | -87% | 1 | 1 | 0% | 3,028 | 663 | -78% | 0 | 0 | — |
case-16 | fail→pass | 17,857 | 2,424 | -86% | 1 | 1 | 0% | 2,116 | 578 | -73% | 0 | 0 | — |
case-17 | fail→pass | 17,482 | 10,236 | -41% | 1 | 1 | 0% | 2,278 | 1,052 | -54% | 0 | 0 | — |
case-18 | pass→pass | 20,385 | 11,025 | -46% | 1 | 1 | 0% | 2,591 | 883 | -66% | 0 | 0 | — |
case-19 | pass→pass | 19,537 | 4,732 | -76% | 1 | 1 | 0% | 1,733 | 905 | -48% | 0 | 0 | — |
case-20 | fail→pass | 104,292 | 3,278 | -97% | 1 | 1 | 0% | 6,362 | 699 | -89% | 0 | 0 | — |
case-21 | pass→pass | 30,501 | 8,593 | -72% | 1 | 1 | 0% | 1,153 | 710 | -38% | 0 | 0 | — |
case-22 | pass→pass | 20,777 | 9,762 | -53% | 1 | 1 | 0% | 1,486 | 739 | -50% | 0 | 0 | — |
case-23 | fail→pass | 38,382 | 4,713 | -88% | 1 | 1 | 0% | 1,996 | 941 | -53% | 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, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +48 percentage points is the difference between those two pass rates over the 22 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 | +45% |
| gemini-3.6-flash | verified | 8/4/2026 | +59% |
Other measured skills in the registry, with their headline benchmark lift.