Summary

This pull request flips all three implementations from assistant_view to agent_view to opt into the new Agent DM Messages Tab experience.

• Rename features.assistant_view → features.agent_view in every manifest, and the inner assistant_description → agent_description.
• Remove assistant_thread_started from settings.event_subscriptions.bot_events.
• Merge the existing set_suggested_prompts logic into the app_home_opened handler, gated on event["tab"]: the Home tab keeps publishing its Block Kit view, the Messages tab pins the suggested prompts.
• Delete the now-unused assistant_thread_started.py handler files.
• Refresh tests/test_app_home_opened.py to cover both event["tab"] branches.
• TODO(agent-dm-messages-tab) markers placed where app_context_changed wiring will land once Bolt for Python exposes the event type. That work is deferred to a follow-up PR.
• README screenshots will be refreshed after the partner flip lands and we can capture the new UX.

Implementation note: Bolt's set_suggested_prompts middleware helper is only injected for assistant_thread_started, so the handler now calls client.assistant_threads_setSuggestedPrompts on the WebClient / AsyncWebClient directly.

Files touched per implementation:

claude-agent-sdk/ (async)

• manifest.json — agent_view rename, drop assistant_thread_started subscription
• listeners/events/app_home_opened.py — branch on event["tab"]
• listeners/events/__init__.py — drop assistant-thread registration
• listeners/events/assistant_thread_started.py — deleted
• tests/test_app_home_opened.py — cover both branches

openai-agents-sdk/ (sync) — same change set
pydantic-ai/ (sync) — same change set

Testing

Setup:

1. cd (e.g. claude-agent-sdk/)
2. Copy .env.sample → .env and fill in ANTHROPIC_API_KEY / OPENAI_API_KEY
3. Install dependencies:
   • JS: npm install
   • Python: pip install -r requirements.txt
4. Run the app: lack run

Functional checks - Messages tab:

5. Open a DM with the app. The DM should render as a traditional Messages tab (not the right-rail Assistant pane).
6. On first open, the three suggested prompts should appear pinned at the top of the Messages tab:
   • starter-agent: "Write a Message" / "Summarize" / "Brainstorm"
   • support-agent (Casey): "Reset Password" / "Request Access" / "Network Issues"
7. Tap a suggested prompt → it should populate the message composer; send it and confirm the agent replies inline.
8. Send a free-form follow-up. The bot should reply in the Messages tab (no thread-only behavior).
9. Send a multi-turn exchange (3+ messages) to confirm conversation history still threads through the agent (per-repo store is unchanged).

Functional checks - Home tab:

10. Click the Home tab in the same DM.
    • starter-agent: should still show the welcome blocks (header + MCP-connected/disconnected section).
    • support-agent (Casey): should still show the 5 category buttons (Hardware / Software / Access / Network / Other).
11. support-agent only: click a category button → the issue submission modal should open. Submit it → confirm a DM is posted and the agent responds.

Cross-cutting checks:

12. Invite the bot to a public channel.
13. @-mention the bot in a public channel → it should reply in the channel thread (no assistant_thread_started regression).
14. Click a thumbs-up / thumbs-down feedback button on any agent reply → confirm no error in the app logs.