Enforce prompt engineering, evaluation, and cost governance for production LLM applications.
npx skills add https://github.com/resultakak/argos --skill llm-ops-resultakak
Please help me install this Agent Skill. Skill: llm-ops Source: https://github.com/resultakak/argos/tree/main/skills/llm-ops Command: npx skills add https://github.com/resultakak/argos --skill llm-ops-resultakak
Production LLM application discipline — prompt engineering (versioned, structured outputs, Anthropic prompt caching) + eval harness (golden set + LLM-as-judge + CI gate) + cost budget (token + cache hit + PR-time delta) + RAG architecture (chunking + embedding + vector DB + reranker + RAGAS eval) + provider abstraction. A model-agnostic core with Anthropic SDK integrated.
Set up an end-to-end LLM workflow with prompt templates, golden-set evaluation, token-budget checks, and RAG-enabled retrieval.
High-intent search queries and answers about installing and using this skill.
Enforce prompt engineering discipline by using versioned prompt templates and structured outputs to ensure consistent LLM behavior across deployments. This standardizes prompt design and maintains behavior consistency during multi-model deployments.
The best way to evaluate LLM applications is integrating an eval harness with golden sets and LLM-as-judge cross-model evaluation. Adding CI gating ensures deployments pass evaluation checks before release, providing cost-aware feedback loops.
Implement RAG architecture by configuring data chunking, embedding generation, vector DB storage, and reranker steps. You evaluate retrieval quality using RAGAS evaluation integrated within multi-provider workflows to ensure accurate data flow.
Yes, you can use provider abstraction to support multi-model deployments. A model-agnostic core integrates the Anthropic SDK and Anthropic prompt caching, allowing you to manage token budgeting and cache hit deltas across different providers.
Monitor token budgeting and cost governance by tracking token usage, cache hit rates, and PR-time cost deltas. This enforces cost-aware feedback loops and ensures LLM operations remain within defined budget limits during prompt iterations.
rnavarychProvide patterns for AI and LLM integration, including RAG architecture and tool use.
ricableTrace, evaluate, and monitor LLM applications across LangChain, OpenAI, Anthropic, and LlamaIndex workflows.
zhangzhang-111-iDesigns RAG pipelines with chunking, embedding, retrieval, and evaluation strategies.
roanbrasilImplement LLM evaluation patterns with LangSmith, RAGAS, and PromptFoo.
