Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Coordinate multiple specialized agents across parallel work streams with quality gates and conflict resolution.
.claude/skills/a5c-ai-parallel-orchestration/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -60% | 0% |
Coordinate multiple specialized agents across parallel work streams with quality gates and conflict resolution.
database-engineerapi-developerui-developertest-engineerdocumentation-writertasks - Tasks from decomposition phasestreams - Work stream definitionsqualityThreshold - Minimum quality score (default: 80)maxParallel - Maximum parallel streams (default: 5)githubRepo - GitHub repo for progress sync (optional)Agents are dispatched based on stream type:
database -> database-engineerapi -> api-developerui -> ui-developertesting -> test-engineerdocs -> documentation-writerinfrastructure -> architectccpm-parallel-execution.js - Standalone parallel executionccpm-orchestrator.js - Phase 5 of full lifecycle| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 28,168 | 15,904 | -44% | 1 | 1 | 0% | 6,205 | 3,464 | -44% | 0 | 0 | — |
case-02 | fail→pass | 20,075 | 11,859 | -41% | 1 | 1 | 0% | 3,699 | 2,779 | -25% | 0 | 0 | — |
case-03 | fail→pass | 19,195 | 20,045 | +4% | 1 | 1 | 0% | 3,711 | 4,772 | +29% | 0 | 0 | — |
case-04 | fail→fail | 20,590 | 1,872 | -91% | 1 | 1 | 0% | 1,532 | 657 | -57% | 0 | 0 | — |
case-05 | fail→pass | 8,135 | 2,073 | -75% | 1 | 1 | 0% | 1,328 | 737 | -45% | 0 | 0 | — |
case-06 | fail→pass | 9,909 | 1,863 | -81% | 1 | 1 | 0% | 1,596 | 654 | -59% | 0 | 0 | — |
case-07 | fail→pass | 17,050 | 1,554 | -91% | 1 | 1 | 0% | 1,456 | 582 | -60% | 0 | 0 | — |
case-08 | fail→pass | 9,776 | 1,931 | -80% | 1 | 1 | 0% | 1,542 | 668 | -57% | 0 | 0 | — |
case-09 | fail→pass | 5,868 | 1,587 | -73% | 1 | 1 | 0% | 937 | 633 | -32% | 0 | 0 | — |
case-10 | fail→pass | 9,169 | 3,300 | -64% | 1 | 1 | 0% | 1,562 | 838 | -46% | 0 | 0 | — |
case-11 | fail→fail | 7,577 | 2,167 | -71% | 1 | 1 | 0% | 1,195 | 775 | -35% | 0 | 0 | — |
case-12 | fail→fail | 7,899 | 2,564 | -68% | 1 | 1 | 0% | 1,346 | 786 | -42% | 0 | 0 | — |
case-13 | fail→pass | 7,101 | 2,177 | -69% | 1 | 1 | 0% | 971 | 692 | -29% | 0 | 0 | — |
case-14 | fail→pass | 4,446 | 1,815 | -59% | 1 | 1 | 0% | 736 | 699 | -5% | 0 | 0 | — |
case-15 | fail→pass | 5,981 | 2,066 | -65% | 1 | 1 | 0% | 995 | 636 | -36% | 0 | 0 | — |
case-16 | fail→pass | 5,150 | 1,839 | -64% | 1 | 1 | 0% | 854 | 682 | -20% | 0 | 0 | — |
case-17 | fail→pass | 6,127 | 1,725 | -72% | 1 | 1 | 0% | 968 | 687 | -29% | 0 | 0 | — |
case-18 | fail→pass | 7,503 | 1,525 | -80% | 1 | 1 | 0% | 1,222 | 636 | -48% | 0 | 0 | — |
case-19 | fail→pass | 11,552 | 1,688 | -85% | 1 | 1 | 0% | 1,930 | 630 | -67% | 0 | 0 | — |
case-20 | fail→fail | 14,619 | 13,216 | -10% | 1 | 1 | 0% | 2,367 | 2,625 | +11% | 0 | 0 | — |
case-21 | fail→fail | 11,797 | 9,332 | -21% | 1 | 1 | 0% | 1,871 | 2,221 | +19% | 0 | 0 | — |
case-22 | fail→fail | 10,197 | 9,249 | -9% | 1 | 1 | 0% | 1,994 | 2,251 | +13% | 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 +68 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.
Other measured skills in the registry, with their headline benchmark lift.