test-repo/post-restart-checklist.md

📋 Complete Post-Restart Verification Checklist

Created: February 25, 2026 Purpose: Verify all infrastructure fixes work after OpenClaw gateway restart

---

🚨 CRITICAL: OpenClaw Update Handling

After OpenClaw updates, configuration may be reset. Always check and restore:

1. Check openclaw.json agents section:

   grep -A 20 '"agents"' /Users/mattbruce/.openclaw/openclaw.json
   

2. Restore subagent configuration if missing:

   "entries": {
     "alice-researcher": {
       "workspace": "/Volumes/Data/openclaw/workspace/agents/alice-researcher"
     },
     "bob-implementer": {
       "workspace": "/Volumes/Data/openclaw/workspace/agents/bob-implementer"
     },
     "charlie-tester": {
       "workspace": "/Volumes/Data/openclaw/workspace/agents/charlie-tester"
     }
   },
   "allowAny": true
   

3. Verify agent directories exist:

   ls -la /Users/mattbruce/.openclaw/agents/
   # Should show alice-researcher, bob-implementer, charlie-tester
   

4. Test subagent access:

   agents_list  # Should show 4 agents
   sessions_spawn({task: "test", agentId: "alice-researcher"})  # Should work
   

⚠️ This step is REQUIRED after any OpenClaw update to restore subagent functionality.

🔄 Subagent System (Critical Priority)

• agents_list returns: main, alice-researcher, bob-implementer, charlie-tester
• sessions_spawn works: Alice can be spawned with test task
• Subagents can read their SOUL.md and workspace files
• Subagent status cron reports active/idle correctly
• Alice/Bob/Charlie can access their specialized workspaces

🏗️ Composable Architecture (High Priority)

• Skills use modular components (extract-content, create-document, etc.)
• No direct Supabase calls in CLI scripts or skills
• API passthrough pattern: CLI → API → Database
• Mission Control CLI calls Mission Control API (not Gantt Board)
• Skills handle multiple input types: URLs, files, raw data

🌐 API & Endpoint Management

• Gantt Board API: https://gantt-board.vercel.app/api (production)
• Mission Control API: POST /api/documents endpoint exists
• Skills can create documents via API instead of direct Supabase
• Cron jobs use production endpoints (no localhost)
• Programmatic authentication works for all APIs
• API responses include proper error handling

🤖 Enhanced Agent Workflows

• Agents use checklist format for task updates
• Checklists enable perfect resumability
• Status transitions: open → in-progress → review → done
• Progress tracking shows what/when/why for each step
• Agents can resume from last incomplete checklist item

⚙️ Configuration Management

• TOOLS.md contains all endpoints, credentials, variables
• MEMORY.md has complete workflow documentation
• openclaw.json has correct subagent configuration
• No hardcoded localhost URLs in automation
• Configuration changes persist across sessions

🔄 Cron Job Operations

• Gantt Board Task Worker uses production API
• Subagent monitor spawns agents when tasks available
• Authentication works without manual intervention
• Error handling and logging functions correctly
• Tasks are processed in priority order

📚 Research & Document Workflows

• "Document this: [input]" creates Mission Control documents
• "Research this: [input]" follows full analysis workflow
• Skills use checklist progress tracking
• Implementation plans attached for ADOPT/EVALUATE_LATER
• Documents appear in Mission Control with proper metadata

📅 Sprint & Task Management

• Tasks created in correct sprint (auto-detected by date)
• Sprint status updates based on date ranges
• Task worker finds and processes active sprint tasks
• No tasks assigned to completed or planning sprints
• Task comments use structured checklist format

📖 Memory & Documentation

• MEMORY.md contains all workflow patterns
• Lessons learned prevent past mistakes
• Composable architecture principles documented
• Subagent setup process clearly explained
• Configuration troubleshooting guides available

🧪 Integration Testing

• End-to-end research workflow: URL → Document + Task + Analysis
• Subagent delegation: Spawn → Work → Report → Complete
• Cron job automation: Task finding → Agent spawning → Completion
• API consistency: All services use same auth/error patterns
• Cross-project communication: Gantt ↔ Mission Control APIs

---

🚨 Failure Recovery

If subagents don't work:

1. Check /Users/mattbruce/.openclaw/agents/ structure
2. Verify openclaw.json has correct entries + allowAny
3. Restart OpenClaw gateway again
4. Check MEMORY.md subagent setup guide

If APIs don't work:

1. Verify TOOLS.md has correct production URLs
2. Test endpoints manually with curl
3. Check authentication is programmatic
4. Confirm no localhost references

If skills fail:

1. Check they use modular components
2. Verify no direct DB calls
3. Test with different input types
4. Check error handling

---

📊 Success Metrics

• Subagents: 3 active specialized agents spawnable
• APIs: All endpoints working with production URLs
• Workflows: Zero direct DB calls, full composability
• Automation: Cron jobs fully operational
• Documentation: All patterns preserved in MEMORY.md

---

🧪 Quick Test Commands

# Subagents
agents_list
sessions_spawn({task: "test", agentId: "alice-researcher"})

# APIs
curl https://gantt-board.vercel.app/api/sprints
curl http://localhost:3001/api/documents

# CLI
cd /Users/mattbruce/Documents/Projects/OpenClaw/Web/gantt-board
./scripts/task.sh list

# Skills (test with URL)
# Research workflow should use checklists

---

📝 Session Summary

This checklist verifies all fixes from the February 25, 2026 session:

• Subagent accessibility issues resolved
• Mission Control CLI architecture fixed
• Composable skill architecture implemented
• Enhanced agent workflows with checklists
• Configuration management rules established
• Cron job operations with production endpoints

Run this checklist after OpenClaw restart to ensure all systems are operational.