Automations & Agents prompts
Prompts for designing automations in n8n, Make and Zapier, writing system prompts for agents, defining tools and guardrails, testing edge cases and documenting what you built. They push for explicit inputs, outputs and failure handling, which is what separates a demo from something a business relies on.
10 prompts · Hebrew and English · Free to copy
Design an automation before opening n8n
Trigger, steps, data shape, failure handling. On one page, before you drag a node.
System prompt for an agent that stays in its lane
Role, rules, tools, refusals and output format. Everything a reliable agent needs.
Extraction prompt that returns clean JSON every time
Messy text in, fixed fields out, and 'null' instead of guesses.
Define the tools your agent is allowed to use
Names, inputs, outputs and the one rule per tool that prevents damage.
Edge-case test set for an agent or workflow
Thirty inputs designed to break it, each with the behaviour you'd accept.
Fix a knowledge assistant that answers confidently and wrong
Chunking, retrieval threshold, citations and refusals: the four dials that matter.
Document a workflow so someone else can run it
The handover doc: what it does, where it lives, what to do when it breaks.
Grading rubric for an agent's answers
Score outputs consistently, by hand or with a second model, and know when quality drops.
Estimate what an automation will cost to run
Model tokens, API calls, tool tiers: the monthly bill before you promise a client anything.
Where the human belongs in your automation
Decide what runs alone, what gets reviewed, and what always waits for a person.
Other categories
Want a prompt for something specific?
Describe what you need and the generator writes one for you, in your language.