Create new logic unit for interactive browser sessions. Reference:
928
This requirement blocks:
934
930
User Story
As a software automation engineer
I want to deploy an autonomous AI agent that can navigate web sites
and
can take action to complete online tasks
So I can replace undesirable alternatives, specifically:
human interaction (expensive, error-prone)
record-and-playback tools (easily broken on dynamic, JavaScript-heavy sites and when web pages change)
scripting (requires advance knowledge of websites and use cases)
and
So I can create new automations that enhance the user experience
Criteria for User Acceptance Testing
Given that an agent has instructions to accomplish an online task of unknown complexity
When it is given a domain-level URL (i.e., a website, not a specific page)
The agent is able to navigate the site and complete the task without further assistance
and
when there is an error or failure to complete the task, the agent communicates this clearly and logs the results for follow-up
Dev notes: Required functionality
[ ] respond to questions about text / data
[ ] Clicking links
[ ] navigating tabs opened externally
[ ] submitting form information (ie, filling in and submitting)
Additional notes on agent instructions:
Option to provide explicit guardrail instructions, e.g., whether the agent can click links outside the given domain, and if so, how many hops.
Forms: when filling out fields, what data is it allowed and not allowed to fill in. E.g., if you need to enter an email or other PII to get access to a document, what information is it allowed to use and where does it get that information from?
Create new logic unit for interactive browser sessions. Reference:
928
This requirement blocks:
934
930
User Story
Criteria for User Acceptance Testing
The agent is able to navigate the site and complete the task without further assistance and when there is an error or failure to complete the task, the agent communicates this clearly and logs the results for follow-up
Dev notes: Required functionality
Additional notes on agent instructions:
Configuration Overview