Compétences llm-app-patterns
📦

llm-app-patterns

Sûr

Build Production LLM Applications

Building LLM applications requires navigating complex architectural decisions. This skill provides battle-tested patterns for RAG pipelines, agent systems, and production operations.

Prend en charge: Claude Codex Code(CC)
🥉 73 Bronze
1

Télécharger le ZIP du skill

2

Importer dans Claude

Allez dans Paramètres → Capacités → Skills → Importer un skill

3

Activez et commencez Ă  utiliser

Tester

Utilisation de "llm-app-patterns". User asks: What is the company's refund policy?

RĂ©sultat attendu:

  • Retrieves relevant policy documents from vector database
  • Generates answer grounded in retrieved context with source citations
  • Returns response with confidence score and document references

Utilisation de "llm-app-patterns". User asks: Plan a research project on climate change impacts

RĂ©sultat attendu:

  • Creates plan with steps: gather data, analyze trends, identify sources, draft report
  • Executes each step sequentially with tool calls
  • Synthesizes findings into comprehensive research outline

Audit de sécurité

Sûr
v1 • 2/25/2026

This skill is a documentation file containing educational content about LLM application patterns. All static analysis findings are false positives caused by markdown formatting. The backticks flagged are code block delimiters and ASCII art borders, not shell command execution. URLs are documentation references, not active network calls. Code examples like hashlib.sha256 are illustrative and use secure algorithms. No executable code or security risks detected.

1
Fichiers analysés
763
Lignes analysées
0
résultats
1
Total des audits
Aucun problème de sécurité trouvé
Audité par: claude

Score de qualité

38
Architecture
100
Maintenabilité
87
Contenu
50
Communauté
100
Sécurité
83
Conformité aux spécifications

Ce que vous pouvez construire

RAG Knowledge Base

Build a question-answering system grounded in your documentation using hybrid search and contextual compression.

Agent Task Automation

Create multi-step agents that can search, calculate, and synthesize information using the ReAct or Plan-and-Execute patterns.

LLM Production Monitoring

Implement observability for LLM applications with metrics tracking, distributed tracing, and evaluation frameworks.

Essayez ces prompts

Basic RAG Prompt 
Answer the user's question based ONLY on the following context. If the context doesn't contain enough information, say you don't have enough information.

Context:
{context}

Question: {question}

Answer:
ReAct Agent Prompt 
You are an AI assistant that can use tools to answer questions.

Available tools:
{tools_description}

Use this format:
Thought: [your reasoning about what to do next]
Action: [tool_name(arguments)]
Observation: [tool result]
... (repeat as needed)
Thought: I have enough information to answer
Final Answer: [your response]

Question: {question}
Few-Shot Prompt with Examples 
Input: {example1_input}
Output: {example1_output}

Input: {example2_input}
Output: {example2_output}

Input: {user_input}
Output:
Prompt Chain for Research 
Step 1 (Research): Research the topic: {input}
Step 2 (Analyze): Analyze these findings: {research}
Step 3 (Summarize): Summarize this analysis in 3 bullet points: {analysis}

Bonnes pratiques

  • Use hybrid search combining semantic and keyword matching for better retrieval accuracy
  • Implement caching for deterministic prompts to reduce latency and costs
  • Track key metrics like latency, token usage, and user satisfaction for continuous improvement

Éviter

  • Using fixed-size chunking without considering document structure, which breaks context
  • Skipping evaluation and monitoring, making it impossible to detect quality degradation
  • Not implementing fallback strategies when primary LLM providers experience outages

Foire aux questions

What is the difference between RAG and fine-tuning?
RAG retrieves relevant documents at query time and provides them as context, allowing the model to access up-to-date information without retraining. Fine-tuning adjusts model weights on training data, which is better for learning style or format but cannot add new knowledge after training.
How do I choose between different agent architectures?
Use ReAct for interactive multi-step reasoning with tool use. Use function calling for structured API integrations. Use Plan-and-Execute for complex tasks requiring upfront planning. Use multi-agent systems when specialized expertise is needed for different subtasks.
What vector database should I use for production?
For managed service with high scale, use Pinecone. For self-hosted with GraphQL, use Weaviate. For existing PostgreSQL infrastructure, use pgvector. For prototyping, use ChromaDB for simplicity.
How can I reduce LLM API costs?
Implement semantic caching for similar queries, use smaller models for simple tasks, optimize prompt length, set appropriate temperature (0 for cacheable responses), and use fallback models when primary is rate-limited.
What metrics should I track for LLM applications?
Track performance metrics (latency, tokens per second), quality metrics (user satisfaction, task completion rate), cost metrics (cost per request, cache hit rate), and reliability metrics (error rate, timeout rate).
How do I handle hallucinations in RAG systems?
Instruct the model to say 'I don't know' when context is insufficient, use groundedness scoring in evaluation, implement retrieval quality checks, and consider adding a verification step where the model cites sources for claims.

Détails du développeur

Auteur

sickn33

Licence

MIT

DĂ©pĂ´t

https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/llm-app-patterns

RĂ©f

main

Structure de fichiers

đź“„ SKILL.md