A technical guide to prompting Agents on Public

How the AI works

Most AI tools operate linearly: you enter a prompt, and the system generates an output based on the information available.

Agents on Public are designed differently. When you describe a strategy in plain English, the AI doesn’t immediately produce an active Agent. Instead, it evaluates your instructions and asks follow-up questions if key parameters are missing.

If your initial prompt is broad or unclear—for example, “Buy the dip on tech stocks”—the system will request clarification. Which specific assets? What percentage decline qualifies as a dip? Which account should execute the trade? How much capital should be allocated? Should execution occur immediately or at a defined time?

This interactive process is deliberate. In trading and investing, the details matter—so your Public Agent will make sure to get all the details right before proposing a workflow.

This interactive process is deliberate. We want the AI to clarify your intent, lock in the specifics, and convert your idea into a precise, executable workflow before anything is activated.

How to think about prompting

At the heart of every Agent is a simple idea: you define a task, and the AI figures out how to make it happen. You don’t need to think in code, or decision trees, or formal rules. You just need to describe your intent—clearly enough that the AI can ask the right follow-up questions and translate it into a precise, executable workflow.

Expressing your intent

Describe the trigger and the action:
The most direct way to prompt an Agent is to describe a specific condition and what you want to happen when it’s met. The AI will work with you to fill in any missing parameters.

Add a broader goal:
For more sophisticated workflows, you can use a precise trigger to activate an Agent that then pursues a broader objective—letting the AI handle the execution complexity.

Set boundaries and parameters:
When building an Agent workflow, our AI will always work with you to set clear boundaries for your Agent to operate within. You don’t have to express these in your initial prompt, but you can use them to dictate the size and scope of the action.

Boundaries might include a strict capital allocation (e.g., “Invest exactly $5,000”), an asset-level risk parameter (e.g., “Stop out if AAPL drops more than 10% from my entry price”), or the frequency an agent executes its goal (e.g., “Check every 5 minutes but only allow up to 2 purchases a day)

You always have the final say

Before any Agent goes live, you’ll have the opportunity to review its complete workflow—every condition, every action, every parameter—and make changes with follow-up prompts until the logic looks exactly right. Be sure to review both the description and workflow before activating it. Nothing executes until you approve it.

Once an Agent is active, you retain full control. You can revisit its logic at any time, adjust conditions, change capital allocations, pause it temporarily, or shut it down entirely. Your Activity Feed gives you a running log of everything your Agent evaluates and acts on, so you always know exactly what’s happening in your account.

Agent skills and capabilities

Agents will be able to work across five broad areas of your portfolio. Here’s a sense of the range—and the kinds of conditions each one can respond to.

Trading Strategies

From one-time buys to multi-leg options strategies, Agents can execute across stocks, options, and bonds based on conditions you define. Triggers can range from a price threshold to an indicator or volatility spike at a set time of day.

Cash Management

Agents can monitor your account balances and automatically move cash—sweeping idle funds into yield accounts, maintaining minimum liquidity thresholds, or reallocating when opportunities arise.

Reading Indicators

Agents have the ability to read and interpret key technical financial indicators, including EMA, SMA, SMI, Bollinger Bands, and more. This enables data-driven analysis of market trends, momentum, and volatility to support informed trading and investment decisions.

Accessing Data Sources

Agents have access to Public’s real-time data feeds. Equipping the agent with access to a wide range of financial data sources, enabling comprehensive, multi-dimensional market research by grounding analysis in real, up-to-date financial data across asset classes and timeframes.

Watch the keynote demo to learn more: