Problem / Motivation
When a sub-agent is delegated part of a task, the interface only shows a single collapsed entry with a status badge and a duration — there is no way to see what that agent is actually doing while it works, or what it did after it finishes. Users have to wait for a final result with no visibility into the agent's own steps, reasoning, or tool usage along the way.
Goal
Give users a way to look inside a running or completed agent and follow its own activity — the steps it takes, the actions it performs, and how it arrives at its result — instead of only seeing a single opaque status line.
Scope
- A way to open a detail view for any agent shown in the agents list, while it is running or after it has finished.
- The detail view shows the agent's own sequence of actions/steps as they happen (live updates while running), not just a final summary.
- Clear indication of the agent's current status and progress within its own work, not just "running" vs "done".
- Errors or failures inside the agent's own work are visible in this detail view, not just a generic failure badge.
Non-goals
- Changing how top-level (non-agent) messages or tool calls are displayed.
- Building this for arbitrary third-party agent runtimes that don't expose their internal steps — scope is limited to agents launched from within this app.
Acceptance Criteria
Open Questions
- Should this level of detail be available for every kind of agent delegation, or only for larger/longer-running ones where it adds the most value?
Problem / Motivation
When a sub-agent is delegated part of a task, the interface only shows a single collapsed entry with a status badge and a duration — there is no way to see what that agent is actually doing while it works, or what it did after it finishes. Users have to wait for a final result with no visibility into the agent's own steps, reasoning, or tool usage along the way.
Goal
Give users a way to look inside a running or completed agent and follow its own activity — the steps it takes, the actions it performs, and how it arrives at its result — instead of only seeing a single opaque status line.
Scope
Non-goals
Acceptance Criteria
Open Questions