Building with AI prompts
System prompts, evals, agent tool design and RAG debugging for teams shipping AI products.
Write a production system promptTop pick
Drafts a system prompt for an AI feature: role, context, rules, tools, output format and examples.
Build an eval setTop pick
Creates test cases and grading criteria for an LLM feature, so prompt changes are measured, not guessed.
Specifies the tools an agent needs: names, descriptions, parameters and failure modes the model can use well.
Finds whether bad answers come from retrieval, chunking, ranking or generation, with tests for each.
Rewrites a weak prompt with clear structure, examples and an output format, and explains every change.