QA: Sub-Agents — Mastra

Prerequisites

• Demo is deployed and accessible
• Agent backend is healthy (check /api/health)

Test Steps

1. Basic Functionality

• Navigate to the subagents demo page.
• Verify the delegation log card loads (data-testid="delegation-log").
• Verify the empty-state message renders: "Ask the supervisor to complete a task. Every sub-agent it calls will appear here."
• Verify the delegation count (data-testid="delegation-count") reads "0 calls".
• Verify the chat panel input placeholder is "Give the supervisor a task...".

2. Sub-agent delegation flow

• Click the "Write a blog post" suggestion. The supervisor should run a research -> write -> critique sequence.
• Verify a "Supervisor running" (data-testid="supervisor-running") badge appears while the run is active.
• Verify delegation entries (data-testid="delegation-entry") stream in as the supervisor calls each sub-agent.
• Verify the entries are tagged with the correct sub-agent label: Research / Writing / Critique, with their respective emoji (🔎 / ✍️ / 🧐).
• Verify each entry shows: - a "#1", "#2", "#3" index - the task text the supervisor passed in - the sub-agent's text result in the white inner panel - a "completed" status pill
• Verify the delegation count updates to match the number of entries.

3. Subsequent runs accumulate

• After the first run finishes, click "Explain a topic". Verify new delegation entries are APPENDED to the log (not replacing the previous run's entries).
• Verify entries from both runs remain visible.

4. Ad-hoc delegation

• Send a freeform message: "Just research the topic of LLM tokenization, no draft needed."
• Verify only a Research delegation entry appears (the supervisor is not forced to run all three).

5. Suggestions

• Verify "Write a blog post", "Explain a topic", and "Summarize a topic" suggestion buttons are visible.

6. Error Handling

• Send an empty message — should be ignored or handled gracefully.
• Verify no console errors during normal usage.
• If a sub-agent errors, the delegation entry should still appear with a clear error result string (prefixed with [sub-agent error]).

Expected Results

• Delegation log loads within 3 seconds.
• Supervisor responds within 30 seconds for the canonical research -> write -> critique flow (slower than other demos because three nested LLM calls run sequentially).
• Each delegation entry surfaces in the log without a page refresh.
• The delegation count matches the number of entries shown.
• No UI errors or broken layouts.