Prompts
AI agent prompts and templates
Free, reusable prompts for AI agents and the model calls they rely on. Choose a template for tool calling, RAG, extraction, or support, then adapt its variables and test the example in your app.
New to agent instructions? Start with how to write an AI agent prompt. For installable coding procedures, browse agent skills.
AI agent system prompt template
A reusable agent system prompt with an objective, tool boundaries, stopping conditions, and an evidence-based final report.
Agents · intermediateAnswer with citations
Write an answer where each claim points at a source passage.
RAG · intermediateContent moderation classifier
Flag a message against a short policy, and quote the part that triggered it.
Moderation · intermediateCustomer support answer
Draft a support reply from a policy passage and the fields triage already extracted.
Support · intermediateDocument classifier
Assign a document to one type so later steps can use the right schema.
Classification · beginnerEntity extractor
List the people, organizations, products, and dates a document actually names.
Extraction · beginnerIntent classifier
Choose one label from a fixed list for a user message.
Classification · beginnerJSON output system
Instructions that keep a model on a schema and silent about everything else.
Structured output · beginnerLanguage translator
Translate text and leave glossary terms in the form you specify.
Transformation · beginnerMeeting action items
Pull owners, tasks, and due dates out of a transcript as data.
Summarization · intermediateMeeting summarizer
Brief someone who missed a meeting, without adding decisions that were not made.
Summarization · beginnerNatural language filter parser
Turn a sentence about filters into a structured query your UI can apply.
Search · intermediatePrompt evaluator
Compare a prompt change against saved cases and say what got better or worse.
Evaluation · advancedRAG context compressor
Shorten retrieved passages without dropping the sentences that answer the question.
RAG · intermediateRAG grounded answer
Answer a question using only the passages that were retrieved.
RAG · intermediateResponse quality grader
Score one answer against a rubric and quote the failure.
Evaluation · advancedSafe tool calling system
A system prompt that lets a model request tools and refuses anything outside that list.
Agents · advancedSearch query rewriter
Turn a conversational question into a short search query for your documents.
Search · intermediateStructured data extractor
Extract a JSON object from messy user text.
Extraction · beginnerSupport ticket triage
Label a support ticket, pull out the facts a reply needs, and name the next queue.
Support · intermediateTool selection router
Pick the one tool that should handle a request, or say that none apply.
Tool calling · advancedChoose a prompt for the job
An agent prompt describes a task, the context to use, available tools, and when to stop. A prompt guides the model; your application must still enforce permissions, validate tool arguments, and limit retries.
How to use these templates
- Open a template and check its intended use and example output.
- Replace every variable with your actual tools, context, or schema.
- Keep stable instructions separate from untrusted user input and retrieved content.
- Test a successful task, missing information, and a failed tool call before shipping.
Use the evaluation guide to compare revisions against the same cases.
What is the difference between a prompt and a skill?
A prompt supplies instructions for a model call. A skill packages a repeatable procedure for a coding agent, including steps and checks. Browse coding agent skills for installation-ready procedures, or implementation guides to build a complete feature.