codexskills.dev

AI Engineering

Ship LLM features you can measure. Skills for building on models with the discipline of normal software: typed tool calls and structured outputs, retrieval pipelines with measured quality, evaluation sets that gate releases, defenses against prompt injection, and tracing that shows cost and failures.

  • 10 skills
  • 1,708 lines of instructions
  • MIT

What's inside

Best for: Developers adding LLM features to a product. Start with openai-responses-api and llm-evals. Each skill loads when a request matches its description, or call one directly with its $name.

  1. OpenAI Responses API Typed output items, deliberate state, clean streams AI & LLM Engineering$openai-responses-api181 lines
  2. OpenAI Agents SDK Few agents, bounded runs, guardrails that block AI & LLM Engineering$openai-agents-sdk207 lines
  3. OpenAI Structured Outputs Typed JSON from the model, every time AI & LLM Engineering$openai-structured-outputs176 lines
  4. OpenAI Function Calling Tool loops that finish and fail gracefully AI & LLM Engineering$openai-function-calling168 lines
  5. RAG Pipelines Grounded answers with citations that resolve AI & LLM Engineering$rag-pipelines154 lines
  6. RAG Evaluation Know whether retrieval or generation failed AI & LLM Engineering$rag-evaluation145 lines
  7. LLM Evals Ship prompt changes on numbers, not vibes AI & LLM Engineering$llm-evals161 lines
  8. Prompt Injection Defense Assume the text is hostile; make it not matter AI & LLM Engineering$prompt-injection-defense193 lines
  9. LLM Observability Trace any answer back to prompt, context and cost AI & LLM Engineering$llm-observability184 lines
  10. Prompt Caching and Cost Lower the LLM bill without lowering quality AI & LLM Engineering$prompt-caching-cost139 lines
esc