CEO welcome & user agency:
- Removed auto-posted user message — CEO greets the board first
- "CEO is waking up..." → "CEO is composing..." → welcome message fades in
- Response chips ("Yes, get started!" / "Let's discuss first") fade in after
- Planning task created unassigned — CEO only wakes when user initiates
- "Yes" chip sends directly; "Discuss" pre-fills input for editing
Merged steps 5+6:
- "Approve & hire" on the Plan step creates hire tasks directly
- No redundant confirmation step
- Step 6 is now a welcome/orientation screen (hidden from nav tabs)
Orientation screen:
- Shows what to expect: Tasks, Agents, Approvals, Dashboard
- "Go to dashboard" button closes wizard and navigates
Nav tabs cleaned up:
- Mission → Launch → CEO → Plan → Review (5 visible tabs)
- Orientation step exists but not in nav (no redundant rockets)
Chat UX polish:
- Single-line input (like a URL bar) instead of multi-line textarea
- CEO welcome message always visible in conversation history
- Structured task description tells CEO exact format for role specs
- "Confirm mission" button hidden until user has chosen a path
- Switching paths preserves previously entered text
Parser improvements:
- Handles N. **Role Name** with indented bullets (fallback format)
- Summary field populated from first expertise line when no explicit summary
- Better skip patterns for non-role sections
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Chat fixes:
- Comment order: sort chronologically (oldest first)
- Reopen+interrupt: user messages reassign task to CEO so it always wakes up
- Strip markdown links from CEO messages to keep user focused on wizard
- Cycling status messages (rotate every 5s) with elapsed timer
- "Review plan" CTA properly disappears when user sends follow-up
- Fetch plan document (not comment summary) for richer role data
Structured role cards:
- 7 fields: Summary, Expertise, Priorities, Boundaries, Tools, Communication, Collaboration
- Collapsible card view with "Show more" / "Show less"
- Full edit mode with labeled textareas per field
- Hire tasks include structured role spec in description
Plan parser:
- Handles "## Role N: Name" format with ### sub-sections
- Handles "### N. Name" format with **Label:** bullets
- Maps CEO's labels (Why→Summary, Responsibilities→Expertise, etc.)
- Skips non-role sections (Summary, Next Steps, Mission, etc.)
Other:
- localStorage persistence for wizard state (survives page refresh)
- Cleaned up step 6 summary (removed redundant company/CEO entries)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Creates a new OnboardingChat component that provides an embedded chat
experience between the user and the CEO agent during onboarding.
How it works:
- After CEO creation (step 3), a planning task is auto-created and
assigned to the CEO with the company mission as context
- An initial comment kicks off the conversation asking the CEO to
propose a hiring plan
- OnboardingChat polls issuesApi.listComments every 4 seconds
- Messages render as chat bubbles (user on right, agent on left)
- A "thinking" indicator shows when waiting for the agent
- Automatically detects hiring plan patterns in agent responses
(markdown headers/lists with role names)
- Calls onPlanDetected callback when a plan is found
The existing issue comment system is the backbone — no new server
endpoints needed. The agent wakes up automatically when comments
are posted via the existing wakeup-on-comment pipeline.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
