Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Expert in LangGraph - the production-grade framework for building stateful, multi-actor AI applications. Covers graph construction, state management, cycles and branches, persistence with checkpointers, human-in-the-loop patterns, and the ReAct agent pattern. Used in production at LinkedIn, Uber, and 400+ companies. This is LangChain's recommended approach for building agents. Use when: langgraph, langchain agent, stateful agent, agent graph, react agent.
.claude/skills/dokhacgiakhoa-langgraph/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | 55% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -16% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 1% | 0% |
Role: LangGraph Agent Architect
You are an expert in building production-grade AI agents with LangGraph. You understand that agents need explicit structure - graphs make the flow visible and debuggable. You design state carefully, use reducers appropriately, and always consider persistence for production. You know when cycles are needed and how to prevent infinite loops.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 4,489 | 24,644 | +449% | 1 | 1 | 0% | 873 | 1,349 | +55% | 0 | 0 | — |
case-02 | pass→pass | 12,816 | 7,170 | -44% | 1 | 1 | 0% | 1,988 | 1,665 | -16% | 0 | 0 | — |
case-03 | pass→pass | 16,669 | 18,679 | +12% | 1 | 1 | 0% | 3,337 | 3,378 | +1% | 0 | 0 | — |
case-04 | pass→pass | 12,464 | 8,256 | -34% | 1 | 1 | 0% | 2,087 | 1,875 | -10% | 0 | 0 | — |
case-05 | pass→pass | 8,277 | 7,060 | -15% | 1 | 1 | 0% | 1,620 | 1,666 | +3% | 0 | 0 | — |
case-06 | pass→pass | 15,194 | 10,211 | -33% | 1 | 1 | 0% | 2,423 | 2,485 | +3% | 0 | 0 | — |
case-07 | pass→pass | 12,038 | 11,771 | -2% | 1 | 1 | 0% | 1,900 | 2,490 | +31% | 0 | 0 | — |
case-08 | pass→pass | 5,804 | 8,010 | +38% | 1 | 1 | 0% | 1,069 | 1,899 | +78% | 0 | 0 | — |
case-09 | pass→pass | 10,740 | 9,873 | -8% | 1 | 1 | 0% | 1,787 | 2,187 | +22% | 0 | 0 | — |
case-10 | pass→pass | 12,847 | 14,636 | +14% | 1 | 1 | 0% | 2,427 | 2,997 | +23% | 0 | 0 | — |
case-11 | pass→pass | 10,950 | 9,224 | -16% | 1 | 1 | 0% | 1,894 | 2,004 | +6% | 0 | 0 | — |
case-12 | pass→pass | 15,908 | 13,808 | -13% | 1 | 1 | 0% | 2,400 | 2,546 | +6% | 0 | 0 | — |
case-13 | pass→pass | 8,819 | 9,008 | +2% | 1 | 1 | 0% | 1,715 | 1,628 | -5% | 0 | 0 | — |
case-14 | pass→pass | 11,654 | 6,492 | -44% | 1 | 1 | 0% | 1,924 | 1,742 | -9% | 0 | 0 | — |
case-15 | pass→pass | 15,249 | 18,302 | +20% | 1 | 1 | 0% | 2,816 | 2,983 | +6% | 0 | 0 | — |
case-16 | pass→pass | 8,621 | 7,180 | -17% | 1 | 1 | 0% | 1,372 | 1,420 | +3% | 0 | 0 | — |
case-17 | pass→pass | 9,802 | 4,664 | -52% | 1 | 1 | 0% | 1,424 | 1,125 | -21% | 0 | 0 | — |
case-18 | pass→pass | 12,359 | 9,563 | -23% | 1 | 1 | 0% | 1,933 | 1,752 | -9% | 0 | 0 | — |
case-19 | fail→pass | 18,481 | 12,641 | -32% | 1 | 1 | 0% | 3,189 | 2,654 | -17% | 0 | 0 | — |
case-20 | pass→pass | 8,325 | 8,084 | -3% | 1 | 1 | 0% | 1,648 | 1,541 | -6% | 0 | 0 | — |
case-21 | pass→pass | 10,596 | 11,718 | +11% | 1 | 1 | 0% | 1,744 | 2,054 | +18% | 0 | 0 | — |
case-22 | fail→pass | 12,862 | 14,143 | +10% | 1 | 1 | 0% | 2,180 | 2,998 | +38% | 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 +9 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.