Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Multi-agent coordination, critical path method, dependency DAG, and agent allocation optimization
.claude/skills/swarm-optimization-patterns/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | — | — |
| case-16 | ✗→✓ | ▲ Improved | — | — |
| case-07 | ✗→✓ | ▲ Improved | — | — |
| case-06 | ✓→✓ | = Same ✓ | — | — |
| case-20 | ✗→✗ | = Same ✗ | — | — |
Task A (3min) ─┐
├── Task D (5min) ── Task F (2min)
Task B (2min) ─┘ │
└── Task E (3min) ── Task G (1min)
Task C (4min) ─────────────────────────────── Task H (2min)
Critical Path: C → H = 6min (longest path)
Zero-slack: C, H (gecikme tüm timeline'ı etkiler)typescriptinterface TaskNode { id: string agent: string estimatedMinutes: number dependencies: string[] // task IDs priority: 'critical' | 'high' | 'medium' | 'low' } // Topological sort ile execution order function buildExecutionPlan(tasks: TaskNode[]): TaskNode[][] { const layers: TaskNode[][] = [] const completed = new Set<string>() while (completed.size < tasks.length) { const ready = tasks.filter(t => !completed.has(t.id) && t.dependencies.every(d => completed.has(d)) ) layers.push(ready) // Bu layer paralel çalışabilir ready.forEach(t => completed.add(t.id)) } return layers }
| Strateji | Ne Zaman | Nasıl | |----------|----------|-------| | Specialization | Uzman agent var | Task → matching agent | | Load Balancing | Eşit workload | Round-robin + capacity | | Priority-Based | Critical path | Zero-slack task'lara öncelik | | Affinity | Context reuse | Aynı dosyaları kullanan task'lar aynı agent'a |
markdown## Bottleneck Tipleri | Tip | Tespit | Çözüm | |-----|--------|-------| | Resource Contention | Aynı dosya birden fazla agent | Sıralı execute veya lock | | QA Queue | Review bekleyen task yığılması | Paralel reviewer | | Dependency Chain | Uzun sıralı bağımlılık | Task decomposition | | Agent Failure | Tekrarlayan fail | Fallback agent, reassign |
BAĞIMSIZ task'lar → PARALEL
- Farklı dosyalarda çalışan
- Birbirine bağımlı olmayan
- Farklı concern'ler (frontend + backend)
BAĞIMLI task'lar → SIRALI
- Aynı dosyada çalışan
- Output → Input ilişkisi olan
- Schema → Code → Test zinciri| Phase | Geçiş Kriteri | |-------|--------------| | Keşif → Geliştirme | Plan onaylandı, task'lar tanımlı | | Geliştirme → Review | Tüm task'lar QA'den geçti | | Review → Düzeltme | Review feedback var | | Düzeltme → Final | Tüm feedback resolved |
Speedup = 1 / ((1-P) + P/N)
P = parallelizable fraction
N = number of agents
Örnek: %80 parallel, 5 agent
Speedup = 1 / (0.2 + 0.8/5) = 1/0.36 = 2.78x| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-23 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-24 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
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. 24 cases were attempted. The headline lift of +13 percentage points is the difference between those two pass rates over the 24 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.