Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Синхронизируй Gemini docs, memories, git и fullrepo. EN: synchronize Gemini docs, memories, git, and fullrepo.
.claude/skills/bilal140202-ry-sync/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -12% | 0% |
Finalize Gemini adapter work into synchronized source, docs, memories, git, GitHub, and fullrepo state.
Use Gemini-native docs and runtime surfaces. Fullrepo is a branch policy, not a Gemini runtime feature.
Use at the end of a completed task or when a Stop/post-task sync hook asks for final synchronization.
browser artifacts.
Git status, memory state, validator output, commit/tag IDs, fullrepo state, and GitHub sync status.
Do not delete branches, force-push, publish fullrepo, or mutate system configs unless policy and owner authorization allow that exact action.
Working tree is clean, memories are current, runtime markers are absent, and fullrepo state matches policy.
Report exact blockers and the command needed to clear each one.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,373 | 9,741 | +53% | 1 | 1 | 0% | 1,134 | 2,151 | +90% | 0 | 0 | — |
case-02 | fail→fail | 1,762 | 2,272 | +29% | 1 | 1 | 0% | 208 | 670 | +222% | 0 | 0 | — |
case-03 | fail→pass | 5,787 | 7,717 | +33% | 1 | 1 | 0% | 1,024 | 1,617 | +58% | 0 | 0 | — |
case-04 | pass→pass | 10,200 | 5,532 | -46% | 1 | 1 | 0% | 1,899 | 1,377 | -27% | 0 | 0 | — |
case-05 | pass→pass | 5,452 | 3,189 | -42% | 1 | 1 | 0% | 1,227 | 882 | -28% | 0 | 0 | — |
case-06 | pass→pass | 4,556 | 3,441 | -24% | 1 | 1 | 0% | 995 | 1,156 | +16% | 0 | 0 | — |
case-07 | fail→pass | 7,946 | 4,052 | -49% | 1 | 1 | 0% | 1,560 | 1,026 | -34% | 0 | 0 | — |
case-08 | fail→pass | 5,866 | 5,170 | -12% | 1 | 1 | 0% | 1,142 | 1,192 | +4% | 0 | 0 | — |
case-09 | pass→pass | 4,818 | 2,414 | -50% | 1 | 1 | 0% | 813 | 783 | -4% | 0 | 0 | — |
case-10 | fail→pass | 5,310 | 3,279 | -38% | 1 | 1 | 0% | 902 | 887 | -2% | 0 | 0 | — |
case-11 | pass→pass | 7,544 | 3,578 | -53% | 1 | 1 | 0% | 1,433 | 932 | -35% | 0 | 0 | — |
case-12 | fail→pass | 4,019 | 3,611 | -10% | 1 | 1 | 0% | 807 | 708 | -12% | 0 | 0 | — |
case-13 | pass→pass | 6,525 | 3,134 | -52% | 1 | 1 | 0% | 1,167 | 883 | -24% | 0 | 0 | — |
case-22 | fail→pass | 7,444 | 2,438 | -67% | 1 | 1 | 0% | 1,595 | 752 | -53% | 0 | 0 | — |
case-14 | fail→pass | 8,773 | 3,438 | -61% | 1 | 1 | 0% | 1,458 | 831 | -43% | 0 | 0 | — |
case-15 | fail→pass | 12,646 | 3,903 | -69% | 1 | 1 | 0% | 2,181 | 949 | -56% | 0 | 0 | — |
case-16 | fail→pass | 11,784 | 2,006 | -83% | 1 | 1 | 0% | 1,971 | 665 | -66% | 0 | 0 | — |
case-17 | fail→pass | 11,514 | 2,609 | -77% | 1 | 1 | 0% | 1,931 | 736 | -62% | 0 | 0 | — |
case-18 | fail→pass | 7,030 | 2,852 | -59% | 1 | 1 | 0% | 1,366 | 820 | -40% | 0 | 0 | — |
case-19 | fail→pass | 6,197 | 4,187 | -32% | 1 | 1 | 0% | 1,106 | 1,088 | -2% | 0 | 0 | — |
case-20 | fail→pass | 13,572 | 2,032 | -85% | 1 | 1 | 0% | 2,284 | 729 | -68% | 0 | 0 | — |
case-21 | fail→pass | 8,909 | 2,536 | -72% | 1 | 1 | 0% | 1,439 | 741 | -49% | 0 | 0 | — |
case-23 | fail→pass | 10,709 | 3,520 | -67% | 1 | 1 | 0% | 1,774 | 945 | -47% | 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 +65 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.
Other measured skills in the registry, with their headline benchmark lift.