Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Rosetta identifies and routes user request to the most matching workflow
.claude/skills/griddynamics-rosetta/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -64% | 0% |
<rosetta>
<purpose> Routing user request to proper workflow and high process adherence. </purpose>
<prerequisites>
orchestration hitl </prerequisites>
<process>
<most matching workflow>.md (note: "-flow" skills are additional workflows) — YOU MUST FULLY ALWAYS execute loaded workflow following its entire definition for all request sizes, workflow WAS created to fix your failure modes (deviations, and weak process adherence, and shallow analysis), workflow is PRIMARY deterministic process to resolve the original user requestplanning + tech-specs outputs → store per system prompt, never plans/ (read-only)Context loaded using Rosetta: [workflow selected + brief summary], and then execute prerequisites, phase 0, etc as workflow defined, let it drive questioning, planning, execution, review, and validation; no phase skipping;</process>
</rosetta>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 12,137 | 7,310 | -40% | 1 | 1 | 0% | 1,985 | 1,554 | -22% | 0 | 0 | — |
case-02 | fail→fail | 6,333 | 6,657 | +5% | 1 | 1 | 0% | 935 | 879 | -6% | 0 | 0 | — |
case-03 | pass→pass | 6,361 | 4,882 | -23% | 1 | 1 | 0% | 995 | 1,155 | +16% | 0 | 0 | — |
case-04 | fail→pass | 13,403 | 4,736 | -65% | 1 | 1 | 0% | 2,179 | 1,038 | -52% | 0 | 0 | — |
case-05 | pass→pass | 14,548 | 3,030 | -79% | 1 | 1 | 0% | 2,349 | 810 | -66% | 0 | 0 | — |
case-06 | fail→pass | 17,481 | 5,800 | -67% | 1 | 1 | 0% | 2,679 | 1,210 | -55% | 0 | 0 | — |
case-07 | fail→pass | 14,008 | 2,170 | -85% | 1 | 1 | 0% | 2,227 | 588 | -74% | 0 | 0 | — |
case-08 | fail→pass | 12,159 | 2,454 | -80% | 1 | 1 | 0% | 1,835 | 653 | -64% | 0 | 0 | — |
case-09 | pass→pass | 11,237 | 4,435 | -61% | 1 | 1 | 0% | 1,562 | 896 | -43% | 0 | 0 | — |
case-14 | fail→pass | 7,702 | 3,362 | -56% | 1 | 1 | 0% | 1,004 | 846 | -16% | 0 | 0 | — |
case-10 | pass→pass | 10,952 | 6,067 | -45% | 1 | 1 | 0% | 1,612 | 1,132 | -30% | 0 | 0 | — |
case-11 | pass→pass | 14,022 | 2,451 | -83% | 1 | 1 | 0% | 1,936 | 617 | -68% | 0 | 0 | — |
case-12 | fail→pass | 11,478 | 1,820 | -84% | 1 | 1 | 0% | 1,707 | 525 | -69% | 0 | 0 | — |
case-13 | fail→pass | 6,391 | 3,339 | -48% | 1 | 1 | 0% | 1,001 | 800 | -20% | 0 | 0 | — |
case-19 | pass→pass | 14,931 | 14,728 | -1% | 1 | 1 | 0% | 2,299 | 2,530 | +10% | 0 | 0 | — |
case-15 | fail→pass | 11,537 | 4,397 | -62% | 1 | 1 | 0% | 1,737 | 877 | -50% | 0 | 0 | — |
case-16 | pass→pass | 9,794 | 4,072 | -58% | 1 | 1 | 0% | 1,518 | 888 | -42% | 0 | 0 | — |
case-17 | pass→pass | 13,687 | 3,062 | -78% | 1 | 1 | 0% | 1,963 | 864 | -56% | 0 | 0 | — |
case-18 | pass→fail | 6,481 | 6,888 | +6% | 1 | 1 | 0% | 963 | 1,239 | +29% | 0 | 0 | — |
case-20 | pass→pass | 7,424 | 7,641 | +3% | 1 | 1 | 0% | 1,294 | 1,513 | +17% | 0 | 0 | — |
case-21 | pass→pass | 10,778 | 2,962 | -73% | 1 | 1 | 0% | 1,479 | 718 | -51% | 0 | 0 | — |
case-22 | fail→pass | 8,776 | 2,405 | -73% | 1 | 1 | 0% | 1,315 | 644 | -51% | 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. 22 cases were attempted. The headline lift of +41 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.