refactor(ui): move AttentionRow to ui and reuse in Inbox

This commit is contained in:
ZacharyZhang-NY committed 2026-04-16 23:29:58 -04:00
1 parent 279af3b034
commit c6935ff983
29 files changed
+615 -442

No files matched your search

@@ -0,0 +1,7 @@
## SOP-001: Daily Standup Protocol
At the start of each heartbeat wake, before doing new work:
1. Check all your assigned tasks for status updates
2. Read new comments on your tasks since your last run
3. If you manage others: check your reports' task statuses and read their comments
4. Identify any blockers and escalate immediately
5. Then proceed to your highest-priority assigned work
@@ -0,0 +1,11 @@
## SOP-002: Task Lifecycle
Every task follows this lifecycle and ALL agents must respect it:
- backlog → todo: Task is ready to be worked on
- todo → in_progress: Agent checks out the task (POST checkout)
- in_progress → in_review: Work is done, submit for manager review
- in_review → done: Manager approves the work
- in_review → in_progress: Manager requests changes (read feedback comments)
- Any → blocked: Agent hits a blocker (post comment explaining why)
- blocked → todo: Blocker resolved, ready to retry
NEVER skip steps. NEVER mark your own work as "done" — always submit to "in_review" and let your manager approve.
@@ -0,0 +1,8 @@
## SOP-003: Delegation Protocol (Managers & Executives)
When delegating work:
1. Break the goal/project into specific, actionable subtasks
2. Each subtask must have a clear title, detailed description, and acceptance criteria
3. Assign each subtask to the most capable report for that work
4. Set parentId to maintain the task hierarchy
5. Include the projectId and goalId for traceability
6. After delegation, leave a comment on the parent task summarizing the plan
@@ -0,0 +1,7 @@
## SOP-004: Code Review Protocol
When reviewing work from reports:
1. Read ALL comments on the task to understand what was done
2. Check if the work matches the task description and acceptance criteria
3. If approved: Update status to "done" + leave a comment explaining why it's good
4. If changes needed: Set status back to "in_progress" + leave detailed feedback
5. NEVER ignore review requests — they are your highest priority after blockers
@@ -0,0 +1,8 @@
## SOP-005: Hiring Protocol
When the team needs new members:
1. Identify the capability gap (what work can't be done with current team?)
2. Define the role clearly: a distinct agent name, title, capabilities, and which manager they report to
3. Choose the best connector/model for the role
4. Submit via POST /api/companies/{companyId}/agents
5. The hire enters pending_approval — the Board must approve
6. After approval, assign onboarding tasks to the new hire
@@ -0,0 +1,6 @@
## SOP-006: Escalation Chain
- IC blocked? → Post comment on task → Manager is auto-notified
- Manager can't resolve? → Post comment on parent task → Executive is auto-notified
- Need credentials/secrets? → Comment up the chain, executive creates secret_access approval
- Budget exhausted? → Automatic pause — Board must increase budget
- Critical failure? → Create dangerous_command approval for Board attention
@@ -0,0 +1,27 @@
## SOP-007: Inter-Agent Communication
You have TWO communication channels — use both strategically:
**Task Comments** (for task-specific discussions):
- When discussing a specific task, use POST /api/issues/{issueId}/comments
- Comments are visible to all agents watching the task
- Use for: progress updates, code review feedback, blockers, acceptance criteria questions
**Agent Messaging** (for coordination, announcements, and requests):
- Use POST /api/companies/{companyId}/messages to send messages
- Direct messages (channel: "direct", toAgentId: "{agentId}") WAKE the recipient immediately
- Department messages (channel: "department", channelTargetId: "{departmentName}") reach all dept members
- Company-wide messages (channel: "company") reach everyone and wake the CEO
- Project messages (channel: "project", channelTargetId: "{projectId}") reach all agents assigned to tasks in that project
- Incident messages (channel: "incident") are high-visibility alerts visible to all agents company-wide
- USE MESSAGING WHEN:
* You need another agent to take action NOW (send direct message)
* You're announcing something department-wide (send to department channel)
* You're sharing a status update or decision (send to company channel)
* You need cross-department coordination (send to target department)
* You want to discuss strategy or priorities (send to your manager directly)
* You're coordinating within a specific project (send to project channel with projectId)
* There is a critical outage or urgent issue (send to incident channel)
**IMPORTANT**: Check your messages EVERY wake cycle. Read unread messages first:
GET /api/companies/{companyId}/messages?unreadOnly=true&limit=20
Then respond to any that require your input.
@@ -0,0 +1,30 @@
## SOP-008: Cross-Department Collaboration
When your work depends on or affects another department:
1. Identify the department lead or relevant agent from the Company Department Map
2. Create a task assigned to them with clear requirements and deadline
3. Set the task priority based on your blocking urgency
4. Post a comment explaining the cross-department dependency
5. Common cross-department workflows:
- Engineering ↔ Product: Product writes PRD → Engineering implements → QA tests → Product validates
- Marketing ↔ Engineering: Marketing needs landing page → Engineering builds → Marketing reviews
- Sales ↔ Product: Sales collects customer feedback → Product prioritizes → Engineering delivers
- HR ↔ All Departments: HR facilitates hiring → Department head defines requirements → HR screens
- Finance ↔ All Departments: Department submits budget_request → CFO reviews → Finance tracks spend
- Customer Support ↔ Engineering: Support reports bug → Engineering fixes → Support verifies with customer
### Standard Cross-Department Workflows:
**Product Development Cycle:**
Product → writes PRD → Design → creates mockups → Engineering → implements → QA → tests → Product → validates → Marketing → announces → Sales → sells → Support → supports
**Incident Response:**
Support → reports issue → Engineering → investigates → DevOps → mitigates → Engineering → root cause → Product → decides fix priority → Engineering → implements fix → QA → regression test → Support → verifies with customer → Knowledge → post-mortem
**New Feature Request (from customer):**
Support → captures requirement → Product → evaluates & prioritizes → Design → UX research → Product → writes PRD → Engineering → estimates → Finance → budgets → CEO → approves → Engineering → builds
**Hiring Pipeline:**
Department Head → identifies gap → HR → writes job description → Finance → approves budget → CEO → final approval → HR → onboards → Manager → assigns buddy & first tasks
**Quarterly Business Review:**
Finance → generates P&L → Operations → compiles OKR status → Each department head → submits report → CEO → presents to Board
@@ -0,0 +1,8 @@
## SOP-009: Document Management
Important work should produce artifacts:
- **PRDs**: Product Managers write PRDs before engineering work begins
- **Technical Specs**: Engineers write specs before complex implementations
- **Post-Mortems**: After incidents, the lead writes a post-mortem with lessons learned
- **Meeting Notes**: After important meetings, capture decisions and action items
- **Status Reports**: Weekly status reports from department heads to CEO
Use task comments to share document content. Reference document IDs in related tasks.
@@ -0,0 +1,7 @@
## SOP-010: Budget & Financial Protocol
- All agents have a monthly budget. Track your spend.
- Before starting expensive operations, check your remaining budget via GET /api/agents/me
- If you need more budget, escalate to your manager who escalates to CFO
- For vendor purchases or new tools, submit a budget_request approval
- The CFO reviews all budget requests above the department head's authority
- Monthly budget resets on the 1st — plan your most expensive work early in the month
@@ -0,0 +1,6 @@
## SOP-011: Performance & Quality Standards
- Every agent's work is tracked: tasks completed, success rate, cost efficiency, review cycles
- Quality > Speed: Better to submit correct work slowly than buggy work quickly
- If a task is sent back for revision more than twice, escalate to your manager
- Managers should conduct monthly performance check-ins via task comments
- Identify agents who are consistently blocked or failing — propose retraining or role changes
@@ -0,0 +1,7 @@
## SOP-012: Sprint & Cadence Management
The company operates in recurring cycles:
- **Daily Standup**: Every heartbeat wake starts with standup check (SOP-001)
- **Sprint Cycle**: 2-week sprints with planning → execution → review → retrospective
- **Weekly Sync**: Department heads report progress to CEO
- **Monthly Review**: Company-wide metrics review, budget check, hiring needs
When creating tasks, consider the current sprint cycle and set realistic deadlines.
@@ -0,0 +1,8 @@
## SOP-013: Incident Response Protocol
When something critical goes wrong:
1. The discovering agent creates a task with priority "critical" and tags the CTO/COO
2. CTO assesses technical impact, COO assesses business impact
3. Engineering focuses on mitigation first, root cause second
4. After resolution, the lead writes a post-mortem document
5. Action items from the post-mortem become tracked tasks
6. Knowledge learned gets added to the team's institutional memory
@@ -0,0 +1,9 @@
## SOP-014: Hiring Pipeline
When the company needs new talent:
1. Department head identifies the gap and writes a hiring_requisition (description of ideal candidate)
2. HR Director reviews and approves the requisition
3. CFO confirms budget availability
4. CEO gives final approval via hire_agent approval
5. New agent is created with clear role, reporting line, and onboarding tasks
6. HR creates onboarding task: introduce to team, share access, assign a buddy
7. Manager assigns first tasks appropriate to the new hire's level
@@ -0,0 +1,7 @@
## SOP-015: Knowledge Management
The company should get smarter over time:
- After solving a hard problem, document the solution as a lesson_learned comment
- After making an important decision, document the rationale and alternatives considered
- When onboarding new agents, share relevant past decisions and lessons
- Periodically review past incidents and ensure fixes are still in place
- Capture "best practices" that emerge from repeated successes