Unfriendly UX Status message on pentest worker completion

[Yuvanesh-ux]

Description

It looks very odd to have {success, subagentId, findingsCount, objectiveResults} as the status message when a pentest worker completes. This should be changes to something that is more informative in normal english.

Apex version

No response

AI Provider

No response

Steps to reproduce

No response

Screenshots or logs

Operating System

No response

Terminal

No response

[Yuvanesh-ux]

@cursoragent Take a look at this after some deep research implement a solution

[cursor]

Perfect! I've successfully implemented and tested a fix for the unfriendly UX status message issue. Here's a summary of what was done:

Problem Identified

When a pentest worker completed, the entire tool result object {success, subagentId, findingsCount, objectiveResults, message} was being displayed as raw text in the UI, as shown in the screenshot provided.

Root Cause

The AgentMessage component in agent-display.tsx was displaying the message.content field for tool messages, which is always an empty string. The actual tool result (including the user-friendly message field) was stored in message.result but wasn't being extracted and displayed.

Solution Implemented

I modified the AgentMessage component to:

1. Check if a completed tool message has a result object with a message field
2. Extract and display that message for a much better user experience
3. Fall back to the existing behavior for tools without a message field (backward compatible)

Now instead of seeing raw JSON like {success: true, subagentId: "...", findingsCount: 3, ...}, users will see friendly messages like:

• ✓ Worker 'SQLi /api/users' complete. Documented 3 finding(s).

Verification

• ✅ Type check passed (bun run tsc)
• ✅ All tests passed (bun run test)
• ✅ Lint check passed (bun run lint)
• ✅ Build succeeded (bun run build)

PR Created

Pull Request #803 has been created and is ready for review:
#803

The fix is surgical and focused - it only touches the display logic for completed tool messages, ensuring that any tool returning a result with a message field will now show that human-readable message instead of dumping the raw object structure to the UI.