/context output is corrupted/truncated after agent team run completes

Bug Description

After running a task with agent teams (plan mode), when the run completes and I execute /context, the output is corrupted and truncated. Specifically:

1. Overlay corruption: Notification banners (e.g. "Database Authentication Error") render on top of the /context output, cutting through and overlapping the context visualization
2. Truncation: The /context output doesn't show the full result — it gets cut off mid-display. The conversation token summary and overall context usage percentage appears to be missing.

Steps to Reproduce

1. Start a Claude Code session
2. Use agent teams / plan mode to run a complex task (in my case, a project with 20+ agents like security-auditor, typescript-pro, seo-content-auditor, etc.)
3. Wait for the agent team run to complete
4. Run /context
5. Observe that the output is corrupted with overlaid notifications and truncated

Expected Behavior

/context should display the full, clean context visualization showing:

• All context sources (system prompt, conversation, agents, memory files, skills)
• Token counts for each
• Overall context usage percentage
• No visual corruption from other UI elements

Actual Behavior

• Yellow "Database Authentication Error" notification boxes render on top of the /context output (see screenshots)
• The context output appears truncated — agents list and skills list show but the summary is incomplete
• The display is essentially unreadable in parts due to the overlay

Screenshots

<img width="1152" height="2098" alt="Image" src="https://github.com/user-attachments/assets/f5c56244-a7fb-4d6b-8a6a-04911567096d" />

<img width="1152" height="2098" alt="Image" src="https://github.com/user-attachments/assets/4fc43479-1c27-4803-9c78-140ee2153ccc" />

Environment

• Claude Code version: 2.1.34
• OS: macOS (Apple Silicon)
• Terminal: iTerm2
• Model: Opus 4.6
• Context: After team of agent runs