AgentSkillsCN

genai-integration

为应用程序集成 GenAI 模型、工作流以及可观测性功能提供专业指导。（适用于 LLM/Agent/RAG 集成的设计与实施阶段）

SKILL.md
--- frontmatter
name: genai-integration
description: Expert guidance for integrating GenAI models, workflows, and observability into applications. (use when designing or implementing LLM/agent/RAG integrations)
version: 1.0.0

Purpose

Teach the agent how to handle GenAI integration tasks — selecting models, building prompt templates, RAG pipelines, cost optimization, and validation workflows.

When to Apply

Use this Skill when the user asks to:

  • •integrate a GenAI API into an application
  • •design RAG workflows, embeddings pipelines, or agents
  • •build prompt templates or schema-validated prompts
  • •write automation for cost or token optimization
  • •add testing, logging, or observability around GenAI tasks

Instructions

  1. •

    Detect Task Intent

    • •Identify if the request is about GenAI model selection, API integration, workflow design, or optimization.
    • •If the task involves specific frameworks (Node/Python, serverless, Vercel/AWS), include relevant context.
  2. •

    Model & Provider Guidance

    • •Recommend models according to cost, latency, context length, and compliance needs.
    • •Prefer structured outputs (JSON schemas) and function/tool calling where appropriate.
  3. •

    Prompt Engineering

    • •Generate prompt templates: system, developer, and user layers.
    • •Use few-shot examples and explicit output schemas in prompts.
  4. •

    RAG & Embeddings

    • •Break documents into chunks with semantic similarity filtering.
    • •Outline vector store choice and search parameters (faiss/pinecone/weaviate).
  5. •

    Agent Workflows

    • •If task requires agents, design tool use steps, fallback logic, and task decomposition.
    • •Provide stepwise workflows for planning and execution.
  6. •

    Cost & Token Strategy

    • •Suggest caching, batching, model tiering, and token budget limits.
    • •Provide scripts or commands (in scripts/) for automation.
  7. •

    Validation & Safety

    • •Add output validators (schema checks).
    • •Mitigate prompt injection and unsafe operations.

Output Format Guidelines

  • •Include JSON-schema blocks where structured output is required.

Examples (Trigger Patterns)

  • •“Integrate LLM for customer support chatbot with RAG”
  • •“Design GenAI prompt templates for summarization API”
  • •“Automate token cost reduction for GenAI calls”