Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Multiplayer game development principles. Architecture, networking, synchronization.
.claude/skills/lingxling-multiplayer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-01 | ✓→✓ | = Same ✓ | -7% | 0% |
| case-02 | ✓→✓ | = Same ✓ | -6% | 0% |
| case-03 | ✓→✓ | = Same ✓ | -1% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 33% | 0% |
> Networking architecture and synchronization principles.
What type of multiplayer?
│
├── Competitive / Real-time
│ └── Dedicated Server (authoritative)
│
├── Cooperative / Casual
│ └── Host-based (one player is server)
│
├── Turn-based
│ └── Client-server (simple)
│
└── Massive (MMO)
└── Distributed servers| Architecture | Latency | Cost | Security | |--------------|---------|------|----------| | Dedicated | Low | High | Strong | | P2P | Variable | Low | Weak | | Host-based | Medium | Low | Medium |
| Approach | Sync What | Best For | |----------|-----------|----------| | State Sync | Game state | Simple, few objects | | Input Sync | Player inputs | Action games | | Hybrid | Both | Most games |
| Technique | Purpose | |-----------|---------| | Prediction | Client predicts server | | Interpolation | Smooth remote players | | Reconciliation | Fix mispredictions | | Lag compensation | Rewind for hit detection |
| Technique | Savings | |-----------|---------| | Delta compression | Send only changes | | Quantization | Reduce precision | | Priority | Important data first | | Area of interest | Only nearby entities |
| Type | Rate | |------|------| | Position | 20-60 Hz | | Health | On change | | Inventory | On change | | Chat | On send |
Client: "I hit the enemy"
Server: Validate → did projectile actually hit?
→ was player in valid state?
→ was timing possible?| Cheat | Prevention | |-------|------------| | Speed hack | Server validates movement | | Aimbot | Server validates sight line | | Item dupe | Server owns inventory | | Wall hack | Don't send hidden data |
| Factor | Impact | |--------|--------| | Skill | Fair matches | | Latency | Playable connection | | Wait time | Player patience | | Party size | Group play |
| ❌ Don't | ✅ Do | |----------|-------| | Trust the client | Server is authority | | Send everything | Send only necessary | | Ignore latency | Design for 100-200ms | | Sync exact positions | Interpolate/predict |
> Remember: Never trust the client. The server is the source of truth.
This skill is applicable to execute the workflow or actions described in the overview.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 47,532 | 40,514 | -15% | 1 | 1 | 0% | 2,435 | 2,273 | -7% | 0 | 0 | — |
case-02 | pass→pass | 17,591 | 41,455 | +136% | 1 | 1 | 0% | 2,440 | 2,295 | -6% | 0 | 0 | — |
case-03 | pass→pass | 19,160 | 13,693 | -29% | 1 | 1 | 0% | 2,841 | 2,814 | -1% | 0 | 0 | — |
case-04 | pass→pass | 11,522 | 9,949 | -14% | 1 | 1 | 0% | 1,693 | 2,255 | +33% | 0 | 0 | — |
case-05 | pass→pass | 8,887 | 7,066 | -20% | 1 | 1 | 0% | 1,384 | 1,840 | +33% | 0 | 0 | — |
case-06 | pass→pass | 10,231 | 37,085 | +262% | 1 | 1 | 0% | 1,516 | 1,845 | +22% | 0 | 0 | — |
case-07 | pass→pass | 9,637 | 33,777 | +250% | 1 | 1 | 0% | 1,419 | 1,321 | -7% | 0 | 0 | — |
case-08 | pass→pass | 8,483 | 8,599 | +1% | 1 | 1 | 0% | 1,195 | 2,118 | +77% | 0 | 0 | — |
case-09 | pass→pass | 11,631 | 8,381 | -28% | 1 | 1 | 0% | 1,947 | 2,184 | +12% | 0 | 0 | — |
case-10 | pass→pass | 7,304 | 5,263 | -28% | 1 | 1 | 0% | 1,085 | 1,378 | +27% | 0 | 0 | — |
case-11 | pass→pass | 18,040 | 18,891 | +5% | 1 | 1 | 0% | 2,729 | 3,679 | +35% | 0 | 0 | — |
case-12 | pass→pass | 14,007 | 7,121 | -49% | 1 | 1 | 0% | 2,030 | 1,834 | -10% | 0 | 0 | — |
case-13 | pass→pass | 11,784 | 10,578 | -10% | 1 | 1 | 0% | 1,647 | 2,236 | +36% | 0 | 0 | — |
case-14 | pass→pass | 16,232 | 14,733 | -9% | 1 | 1 | 0% | 2,376 | 2,814 | +18% | 0 | 0 | — |
case-15 | pass→pass | 14,670 | 7,640 | -48% | 1 | 1 | 0% | 2,109 | 1,780 | -16% | 0 | 0 | — |
case-16 | pass→pass | 18,909 | 17,355 | -8% | 1 | 1 | 0% | 2,621 | 3,217 | +23% | 0 | 0 | — |
case-17 | fail→pass | 13,068 | 7,180 | -45% | 1 | 1 | 0% | 1,979 | 1,787 | -10% | 0 | 0 | — |
case-18 | pass→pass | 15,296 | 9,245 | -40% | 1 | 1 | 0% | 2,216 | 2,128 | -4% | 0 | 0 | — |
case-19 | pass→pass | 15,636 | 9,895 | -37% | 1 | 1 | 0% | 2,217 | 2,162 | -2% | 0 | 0 | — |
case-20 | pass→pass | 25,228 | 28,718 | +14% | 1 | 1 | 0% | 4,199 | 5,438 | +30% | 0 | 0 | — |
case-21 | pass→pass | 23,819 | 24,000 | +1% | 1 | 1 | 0% | 4,005 | 4,824 | +20% | 0 | 0 | — |
case-22 | pass→pass | 11,751 | 9,111 | -22% | 1 | 1 | 0% | 2,215 | 2,539 | +15% | 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 +5 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.