Prepare a database-change automation
React to selected changed fields while preventing a write-back loop.
Target assistant #
ChatGPT · Required connection: Chatigram plugin.
The public ChatGPT plugin listing is under review. These prompts are for workspaces where the Chatigram plugin is available and connected. Status reviewed 2026-11-10.
What this prompt helps you do #
React to selected changed fields while preventing a write-back loop.
This request checks your organization’s actual capabilities instead of assuming a provider, plan, or node is available.
Customize before using #
Replace the bracketed placeholders with your organization and business details. Keep passwords, API keys, bot tokens, and private customer data out of the prompt. The Chatigram plugin currently requires an Admin role in the selected organization.
Copy the prompt #
Prepare a workflow triggered by [created/updated/deleted] rows in [database/table]. Inspect the real table and choose only the operations we need. For updates, use changed-field information to ignore unrelated changes. The intended outcome is [describe outcome]. If the workflow would write back to the same table, include a guard against repeated self-triggering. Show the trigger, conditions, planned writes, and error behavior. Do not change existing data or enable the workflow.
Use the Chatigram plugin for organization [organization name]. Discover the current permissions, plan limits, triggers, node types, and compatible connections before designing the solution. For conversation workflows, check the selected Line’s Private AI mode before proposing Inbox follow-up or a human handoff. Do not promise human access to private AI conversations. Prepare and save the requested workflows disabled, after showing the draft preview and any unresolved setup. Do not activate, run, assign to a live Line, send messages, or use paid processing. If a paid capability or connection is needed, explain the service, account, transferred data, and required spending approval before adding or binding it. Do not ask me to paste credentials into chat; use protected setup links. Keep unrelated resources unchanged.
Finish with: what was prepared, what is still missing, which plan-dependent features are involved, and the next safe test. Do not claim execution or delivery from a draft preview.
Review the actual result #
Check the preview, branch behavior, required accounts, plan-dependent resources, and any missing setup. A prepared definition is not proof of live execution. Give a later, explicit request for activation, assignment, sends, or paid processing only when you intend those actions.
Return to the AI prompt library for more ideas and target-specific guidance.