Crafting Effective Prompts

The best prompts tell Kindgi the complete story — what triggers the workflow, what data it works with, what it does step by step, and what the end result looks like. Think of it as explaining the task to a new colleague: give them the full picture so they can work independently.

Don't say "my spreadsheet" — say Google Sheet called 'Q1 Leads'. Don't say "send a message" — say post to the #team-updates Slack channel. The more specific you are about which apps, which documents, and which fields, the more accurate your agent will be on the first build.

When your workflow writes to a Google Sheet, reads from an email inbox, or creates cards on a board — list the exact structure. Sheet headers, board names, list names, email subject filters, and status values give Kindgi a precise map to follow. Without them, the agent has to guess, and guesses lead to mismatched columns and extra cleanup.

If your workflow should only act on certain items, spell out the conditions. For example, instead of saying "active leads", write "rows where the Status column equals 'Active'". Explicit filters ensure your agent processes exactly the right data.

Tell Kindgi exactly what the output should look like. If you want an email, describe the subject line, body content, and who receives it. If you want a Trello card, specify the board, list, title format, and which fields go in the description. Detailed output specs = fewer surprises.

When refining your agent in the Studio chat, focus on a single change per message. Asking for three changes at once can lead to unexpected interactions. One clear instruction gives Kindgi the best chance of getting it exactly right.

Start your Studio chat instructions with a clear verb — add, remove, change, move, split, combine. This tells Kindgi exactly what operation to perform instead of leaving it to interpretation.

You can click any step in the Build tab to select it before typing your update. This tells Kindgi exactly which step you're targeting. Pair that selection with a clear description of what to change, and the update will be spot-on every time.

When your agent needs to repeat an action for multiple items (e.g. send an email for each row, create a card for each lead), say so explicitly. Mention what the list is, what to do with each item, and whether anything should happen after all items are processed.

Here are proven prompt structures you can copy and adapt for common automation patterns. Replace the bracketed placeholders with your own details.

These are the most frequent prompt pitfalls we see. Avoiding them will save you time and produce better agents on the first try.