菜单
Monitor system health, detect errors, implement recovery procedures, restart failed processes, maintain audit logs. Use for system health checks, error recovery, process monitoring, or when user mentions "system status", "errors", "logs".
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | monitor-system |
| description | Monitor system health, detect errors, implement recovery procedures, restart failed processes, maintain audit logs. Use for system health checks, error recovery, process monitoring, or when user mentions "system status", "errors", "logs". |
| user-invocable | true |
| allowed-tools | ["Read","Write","Edit","Bash","Glob","Grep"] |
| model | sonnet |
Version: 1.0 Last Updated: 2026-01-12 Purpose: System health monitoring, error recovery, and comprehensive audit logging
This skill provides autonomous system monitoring and self-healing capabilities for the AI Employee. It ensures 24/7 operation by:
Auto-Activate When:
Keywords: "system status", "health check", "errors", "logs", "monitoring", "restart", "recovery"
1. Run health_monitor.py
├─ Check Watcher processes (PID validation)
├─ Verify MCP server connectivity
├─ Check disk space (>10% free required)
├─ Verify network connectivity
└─ Calculate health score
2. If health score < 80:
├─ Log warning
├─ Identify failing components
└─ Trigger recovery procedures
3. If health score < 50:
├─ Create high-priority alert
├─ Attempt auto-recovery
└─ Escalate to user if recovery fails
4. Update System_Status.md with results
1. Error detected in logs
↓
2. Classify severity using error-catalog.md
↓
3. Select recovery procedure from recovery-procedures.md
↓
4. Attempt recovery (up to 3 times with exponential backoff)
↓
5. If successful: Log recovery, resume operations
If failed: Create escalation alert for user
1. Process PID not found or unresponsive
↓
2. Log incident
↓
3. Attempt restart (watchdog_manager.py)
↓
4. Verify process started successfully
↓
5. If 3 restart attempts fail:
└─ Create critical alert in /Needs_Action/
Vault/
├── Logs/
│ ├── system/
│ │ ├── health_checks.json
│ │ ├── incidents/
│ │ └── archives/
│ ├── actions/
│ │ ├── YYYY-MM-DD.json
│ │ └── archives/
│ ├── financial/
│ │ └── archives/
│ └── security/
│ └── archives/
├── System_Status.md
└── Needs_Action/
└── ALERT_*.md (critical escalations)
Complete taxonomy of all error types with severity classifications and recovery strategies.
Step-by-step procedures for recovering from each error category.
Health scoring system with component weights and thresholds.
JSON schema for all log types with retention policies.
Purpose: Continuous health monitoring Frequency: Every 5 minutes Actions:
Purpose: Classify and recover from errors Actions:
Purpose: Comprehensive audit logging Actions:
Purpose: Process monitoring and restart Actions:
Total: 100 points
| Component | Weight | Check |
|---|---|---|
| Watchers | 30pts | All processes running |
| MCP Servers | 25pts | All servers responding |
| Disk Space | 15pts | >10% free |
| Network | 15pts | API connectivity |
| Logs | 10pts | Writing successfully |
| Vault Access | 5pts | Read/write permissions |
Thresholds:
| Log Type | Retention | Archive | Purpose |
|---|---|---|---|
| Actions | 90 days | Compress, keep 1 year | Audit trail |
| System | 60 days | Compress, keep 6 months | Debugging |
| Errors | 180 days | Compress, keep 2 years | Pattern analysis |
| Security | 365 days | Compress, keep 7 years | Compliance |
| Financial | 7 years | Compress, permanent | Legal requirement |
# Run manual health check
python .claude/skills/monitor-system/scripts/health_monitor.py
# View latest status
cat Vault/System_Status.md
# View today's errors
cat Vault/Logs/system/incidents/$(date +%Y-%m-%d).json
# Search for specific error
grep "API_AUTH_FAILURE" Vault/Logs/system/*.json
# Manually restart a watcher
python .claude/skills/monitor-system/scripts/watchdog_manager.py restart gmail_watcher
This skill is working correctly when:
Dependencies:
Provides to Others:
This skill is specifically for system health and operations, not business intelligence.
Next Steps: Reference the detailed documentation in /reference/ for implementation specifics and use the scripts in /scripts/ for actual monitoring operations.
Maintenance: Review error patterns monthly, update recovery procedures as needed, adjust health thresholds based on experience.
Autonomously audit business and financial performance weekly. Analyzes revenue, expenses, completed tasks, bottlenecks, and generates executive briefing with proactive suggestions. Use when generating weekly report, business audit, or when user mentions "briefing", "ceo report", "performance summary", "weekly audit", "business review".
Integrate with Xero accounting system to sync transactions, categorize expenses, track revenue, generate invoices, and maintain financial records. Use when syncing bank transactions, categorizing expenses, creating invoices, checking financial status, or when user mentions "accounting", "xero", "expenses", "revenue", "invoice", "financial report", "reconcile".
Create structured Plan.md files for complex multi-step tasks. Analyzes task requirements, breaks them down into actionable steps with estimates, identifies dependencies between steps, flags steps requiring human approval, and generates comprehensive plans with success criteria. Use when task requires planning, user asks to "create a plan", "plan this task", "break this down", "what are the steps", or when task has multiple components or dependencies.
Manage human-in-the-loop approval workflow for sensitive actions like payments, email sending, social media posts, and file operations. Creates structured approval requests, monitors approval status in /Approved and /Rejected folders, executes approved actions safely, and maintains complete audit logs. Use when any action requires approval, user mentions "check approvals", "pending approval", "approve this", "what needs approval", or before executing sensitive operations like payments, emails, or social posts.
Process emails from Gmail Watcher, categorize by priority (urgent/normal/low), draft appropriate responses using templates, flag items requiring immediate attention. Use when processing emails, checking inbox, handling Gmail notifications, or user mentions "email", "inbox", "gmail", "messages", "check mail", "respond to email".
Process pending tasks, check Needs_Action folder, complete task requests, move finished work to Done folder, and update activity dashboard. Use when the user asks to "process tasks", "check pending tasks", "handle Needs_Action items", "complete my tasks", "check what needs to be done", or mentions the Needs_Action folder, task processing, or wants to see completed work moved to Done.