AgentsMédio

Step 1: Agent runs until it needs to call tool

porlangchain-ai·langchain-ai· v1.0.0 · atualizado em 2026-04-10

Score

INVOKE THIS SKILL when you need human-in-the-loop approval, custom middleware, or structured output. Covers HumanInTheLoopMiddleware for human approval of dangerous tool calls, creating custom middleware with hooks, Command resume patterns, and structured output with Pydantic/Zod.

langchainmiddlewarehuman-in-the-loophitlagent-controltool-approvalagent-framework

0Stars

0Forks

0Usos

Fork

Documento do Skill

SKILL.mdlangchain-middleware/workflow

1. Define the tools that require human approval.

2. Configure the `HumanInTheLoopMiddleware` with the desired policies for each tool (allowed decisions: approve, edit, reject).

3. Integrate the middleware into the LangChain agent.

4. Run the agent with a checkpointer and a thread ID.

5. If a tool call requires approval, the agent pauses execution.

6. Present the tool call details to a human for review.

7. Based on the human's decision, resume execution with the approved, edited, or rejected action.

Telemetria de Agentes

Execuções

total

Taxa de Sucesso

últimos 30d

Latência Média

0.0s

p50

Alucinação

0.0%

detecção

Tokens Entrada

avg 0/exec

Tokens Saída

avg 0/exec

Uso por Plataforma

Skills Relacionados

Compõe com ←framework-selection

70%

Similar aRemembering Conversations

Árvore do Skill

Step 1: Agent runs until it needs to call tool

langchain-middleware

Fases Cognitivas5

1.SENSE

2.CONTEXTUALIZE

3.HYPOTHESIZE

4.ACT

5.REFLECT

Triggers8

implement human-in-the-loop approval for agent actionspause agent execution for human reviewadd middleware to a LangChain agentrequire human approval before a tool is usededit tool arguments before agent executionreject a tool call with feedbackconfigure different HITL policies for different toolsuse HumanInTheLoopMiddleware in LangChain

Avaliar este Skill

Score Breakdown

⭐Avaliação Humana0%

🤖Sucesso de Agentes0%

🕐Atualidade100%

🔗Saúde de Dependências100%

🕸️Centralidade no Grafo0%

🛡️Segurança50%

CompositeScore = α·Humano + β·Agente + γ·Recência + δ·Deps + ε·Centralidade + ζ·Segurança

Instalação

$ synaptic mcp download langchain-middleware

$ synaptic skills detail langchain-middleware

$ synaptic skills live langchain-middleware

Dependências

langchain langgraph @langchain/langgraph @langchain/core zod

Links

GitHub Repository