Uncanny Agent is good at picking up vague requests, but a little extra context goes a long way. These are the habits that consistently produce the best results.
1. Be specific about the trigger and the destination
Instead of “Send my customers an email”, try:
“When a WooCommerce order is paid for any product in the ‘Coaching’ category, wait 1 hour, then send the customer a ‘Welcome to Coaching’ email.”
Uncanny Agent now knows the trigger (payment received), the filter (a specific category), the timing (1-hour delay) and the action (send email). That’s a one-shot recipe.
2. Mention the plugins you use by name
If you have multiple form plugins, CRMs, or payment processors installed, naming the one you want avoids ambiguity. “Use Gravity Forms”, “send to HubSpot”, “in LearnDash” — these short hints save a follow-up question.
3. Use one conversation per topic
Keep building-a-recipe in one conversation, content-drafting in another, and data-lookup in a third. Uncanny Agent uses the rest of the conversation as context, so a focused chat tends to produce sharper answers than a sprawling one. See Managing your conversations for how to start a new chat.
4. Show, don’t just tell
If you’re trying to recreate something—a layout, an error, an email template—attach a screenshot or file rather than describing it. See sharing files and screenshots for how.
5. Iterate in small steps
You don’t have to nail the final result on the first message. A productive flow is:
- Ask Uncanny Agent for a first draft.
- Look at what it built.
- Tell it what to change (“Make the delay 24 hours instead”, “Add an SMS action after the email”, “The tone is too formal”).
Iterative refinement is faster and more accurate than writing one giant prompt.
6. Ask for review and explanation, not just doing
Uncanny Agent makes a great pair-programmer for automations. Try asking:
- “Before you build it, walk me through the recipe you’d suggest.”
- “Why did you pick that trigger?”
- “What could go wrong with this recipe in production?”
These prompts often surface edge cases (a missing token, a race condition, a filter that should be there) before the recipe is saved.
7. Always verify before going live
Even with great prompts, AI can occasionally make mistakes. Before activating a recipe that affects real customers, run it manually in a test with one of your own accounts, or set its Completions allowed limit to 1 while you verify behaviour. See reviewing what Uncanny Agent did for the details that the Agent surfaces after every task.
8. Use 👍 / 👎 generously
The thumbs buttons at the bottom of every response feed directly into how Uncanny Agent improves. If something was great, mark it. If something was wrong, mark it and add a quick note about what was wrong. This is the most effective way to make Uncanny Agent better for everyone.
This Post Has 0 Comments