squad-skills-ai-engineering
AI engineering skills for LLM-powered applications and agent systems: LLM integration, prompt engineering, agent design and orchestration, tool and MCP integration, context engineering, retrieval-augmented generation, evaluation, testing, security, and observability.
Version 0.2.0, 16 skills.
Install
copilot plugin install ./plugins/ai-engineering
Skills
| Skill | Purpose | Load when |
|---|---|---|
agent-design | Agent vs workflow, loop, stop conditions, memory, human-in-the-loop | Designing an agent |
agent-orchestration | Routing, parallel, handoff, shared state, durability, loop control | Coordinating multiple agents or steps |
ai-cost-latency-control | Controlling cost and latency of LLM features in a TypeScript/NestJS chat backend - budgets per user, tenant, and request, token accounting from... | Estimating or capping spend, reducing response latency, adding retries or fallbacks, or setting up cost monitoring |
ai-evaluation | Datasets, graders, LLM-as-judge, CI regression gates | Measuring AI quality |
ai-observability | Traces, token and cost metrics, redacted logging, feedback | Monitoring AI in production |
ai-security | Prompt injection, least privilege, exfiltration, tenant isolation | Hardening AI features |
ai-testing | Fakes, recorded responses, contract tests, vitest patterns | Testing AI code |
chat-application-architecture | End-to-end architecture of a conversational AI application in TypeScript with NestJS, Prisma, PostgreSQL and a Nuxt client - performance first... | Designing or reviewing the data model, API, performance, or request flow of a chat application |
context-engineering | Context budgeting, compaction, memory tiers, caching | Managing what the model sees |
conversation-memory | Conversation memory for chat applications in TypeScript - assembling the context window from stored messages along the active branch, token... | Building or debugging chat history handling, summaries, saved user memories, or personalisation, or when long conversations degrade |
llm-integration | Provider abstraction, streaming, structured output, retries, cost budgets | Calling an LLM from the backend |
mcp-integration | MCP servers and clients, transports, auth, trust | Using or building MCP servers |
prompt-engineering | Prompt structure, roles, few-shot, output contracts, versioning | Writing or changing prompts |
retrieval-augmented-generation | Chunking, embeddings, hybrid search, reranking, citations, access control | Building retrieval pipelines |
streaming-llm-responses | Streaming LLM responses end to end in TypeScript with NestJS and a Nuxt client - parsing provider SSE chunks, accumulating deltas, assembling... | Implementing or debugging streaming chat output, tool-call streaming, stop buttons, reconnects, or stream tests |
tool-integration | Tool schemas, validation, idempotency, permissions, error results | Exposing tools to a model |
Related plugins
Works well with: backend, shared, platform-engineering, data (search and retrieval). Skills may mention skills from other plugins in a short "See also" note; each skill stands on its own if those plugins are not installed.