Task
You are required to design an Agentic AI solution in UiPath consisting of two agents and one escalation app:
Email Intake Agent
Acts as the entry point of the solution.
Monitors a designated mailbox for new incoming emails.
Extracts relevant details such as:
- Sender address
- Subject
- Body (plain or HTML)
- Any URLs or attachments
Passes the extracted data to the Phishing Classifier Agent for analysis.
If the classification is unclear or confidence is low, it triggers the Escalation App for human analyst review.
Finally, it performs the necessary action (move to quarantine, allow, or report) and sends a polite, consistent response to the user confirming that the email was analyzed and handled appropriately.
Phishing Classifier Agent
Analyzes the email content using Agentic reasoning to identify whether the email is:
- Safe
- Suspicious
- Phishing
Provides a structured result including:
- Classification: Safe / Suspicious / Phishing
- Confidence Score: (0–100)
- Reasoning: Short explanation (e.g., “Spoofed domain”, “Malicious link detected”)
- Recommended Action: Allow / Quarantine / Escalate
If the classification confidence is below a defined threshold (e.g., 75%), the agent leaves the classification blank and sets the recommended action to “Escalate.”
Returns the result to the Email Intake Agent for further processing
Escalation App
Launches automatically when classification confidence is low or unclear.
Displays the original email content to the security analyst.
Allows the analyst to manually:
- Review the email
- Select the final classification (Safe / Suspicious / Phishing)
- Choose the final action (Allow / Quarantine / Report)
Once submitted, the reviewed data is passed back to the Email Intake Agent, which continues the workflow and takes the necessary action.
Expected Output
When a new email is received, the system should:
- Automatically classify and handle clear phishing or safe emails.
- Escalate unclear cases through the Escalation App for human validation.
- Respond to the user with a consistent, professional message confirming that their email was analyzed.
- Maintain a log report recording:
- Email ID
- Classification
- Confidence score
- Final action (Auto / Manual)
- Timestamp
License & Service Enablement:
Agent-Building Reference:
Submission Format:
Important Note
- Participants can submit only once
- Submit your solution by clicking Submit Here!
- Deadline: Submit on or before 7 PM Thursday (06/11/2025)
- The submission instructions are mentioned in the Google Form itself
.png)
