{"id":135,"date":"2026-10-17T14:16:03","date_gmt":"2026-10-17T14:16:03","guid":{"rendered":"http:\/\/plugin-docs.com\/?post_type=doc&p=135"},"modified":"2026-10-18T11:05:03","modified_gmt":"2026-10-18T11:05:03","slug":"write-workflow-prompts","status":"publish","type":"doc","link":"https:\/\/dragwyb.com\/docs\/write-workflow-prompts\/","title":{"rendered":"Write Prompts for Workflow Actions"},"content":{"rendered":"\n

A useful workflow prompt tells the model what to do, which data to use, and what kind of result to return. In AI Agent Workflows, prompt fields can use the editor’s variable controls to insert values from the trigger and earlier workflow nodes.<\/p>\n\n\n\n\n

Screenshot to add: AI node prompt field with the variable picker open<\/p>\n\n\n\n

Write a clear prompt<\/h2>\n\n\n
    \n\n
  1. State one task.<\/strong> Tell the model what result you need instead of combining unrelated requests.<\/li>\n\n\n
  2. Provide the necessary context.<\/strong> Use the variable picker to insert trigger fields or outputs from earlier nodes.<\/li>\n\n\n
  3. Set boundaries.<\/strong> Explain what the model should not change, assume, or include.<\/li>\n\n\n
  4. Describe the expected response.<\/strong> Ask for a concise summary, a list, or a structured result when that matches the next workflow step.<\/li>\n\n\n
  5. Test edge cases.<\/strong> Try normal input, missing values, and unexpected content before using the result in a production action.<\/li>\n\n<\/ol>\n\n\n\n

    Use variables instead of guessing field paths<\/h2>\n\n\n

    Use the variable picker in the prompt field to browse the data available from the trigger and connected nodes. The list depends on the workflow graph and available sample data. Capture or load trigger data when needed, then select the real field instead of typing an unverified path.<\/p>\n\n\n\n

    Configure an AI Agent prompt<\/h2>\n\n\n

    Select the AI Agent node and open Parameters<\/strong>. The Source for Prompt (User Message)<\/strong> control determines where the user message comes from. Choose the chat-trigger source when the agent should read the chatInput<\/code> field from a directly connected Chat Trigger. Choose Define below<\/strong> to write a prompt in the Prompt (User Message)<\/strong> field. When the chat-trigger source is selected, the prompt textarea is hidden.<\/p>\n\n\n\n

    Use the optional System Prompt<\/strong> option for persistent instructions that guide the agent’s behavior. Keep the user task and system-level constraints clear, and test the agent with representative inputs.<\/p>\n\n\n\n

    Review results before acting on them<\/h2>\n\n\n

    A prompt cannot guarantee that a model returns correct or complete information. Inspect the node output and validate any fields used by later actions. If the workflow sends messages, updates records, or calls external services, test with limited data first.<\/p>\n","protected":false},"excerpt":{"rendered":"

    Provide clear task instructions and map the right input context into an AI node.<\/p>\n","protected":false},"author":0,"featured_media":0,"parent":128,"menu_order":0,"template":"","class_list":["post-135","doc","type-doc","status-publish","hentry"],"acf":[],"_links":{"self":[{"href":"https:\/\/dragwyb.com\/docs\/wp-json\/wp\/v2\/doc\/135","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dragwyb.com\/docs\/wp-json\/wp\/v2\/doc"}],"about":[{"href":"https:\/\/dragwyb.com\/docs\/wp-json\/wp\/v2\/types\/doc"}],"version-history":[{"count":2,"href":"https:\/\/dragwyb.com\/docs\/wp-json\/wp\/v2\/doc\/135\/revisions"}],"predecessor-version":[{"id":851,"href":"https:\/\/dragwyb.com\/docs\/wp-json\/wp\/v2\/doc\/135\/revisions\/851"}],"up":[{"embeddable":true,"href":"https:\/\/dragwyb.com\/docs\/wp-json\/wp\/v2\/doc\/128"}],"wp:attachment":[{"href":"https:\/\/dragwyb.com\/docs\/wp-json\/wp\/v2\/media?parent=135"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}