task-detection-thinking（任务检测思维）

Task Detection + Proactive Thinking Skill

Automatically detect task anomalies, intelligently analyze causes, and proactively propose solutions

Trigger Conditions

Task Management File Standards

HEARTBEAT.md - Global Task Board Template

## Global Task Board
| TaskID | Task Name | Status | Progress | Deadline | Last Update | Block Reason |
|--------|-----------|--------|----------|----------|-------------|--------------|
| T001 | Initialize three-layer memory | Done | 100% | 2026-03-10 | 2026-03-10 | None |
| T002 | Integrate vector retrieval | Active | 60% | 2026-03-15 | 2026-03-12 | Waiting for API key |
| T003 | Test proactive thinking | Waiting | 0% | 2026-03-20 | 2026-03-11 | Depends on T002 |

Status Definitions:

WORKING.md - Project-Level Task Rules

All subtasks must include:

• Status: Queue/Active/Waiting/Done
• Progress: Percentage (0-100%)
• Owner: Person responsible
• Dependencies: Task IDs this task depends on

Auto-Detection Rules:

• Active tasks stalled >24h → Auto-marked as "Abnormal" 🔴
• Blocked tasks must fill: "Block Reason" + "Support Needed"
• Daily 23:00 → Auto-summary to HEARTBEAT.md

Detection Rules (Auto-execute)

1. Scan HEARTBEAT.md

2. Scan WORKING.md

Proactive Thinking Logic

For "Stalled Tasks"

Analysis Steps:

1. Check last update content → Determine stall type
2. Scan related memories → Find context
3. Generate 3 advancement solutions

Auto-attempt (No human needed):

• Supplement missing info (extract from memory)
• Retry failed interfaces/commands
• Update task status to Waiting (if needed)

Human intervention required:

• Generate detailed reminder, push via Feishu

For "Blocked Tasks"

Analysis Steps:

1. Identify block type (resource/dependency/decision/external)
2. Determine if auto-solvable
3. Record solution to warm memory

Auto-resolve scenarios:

• Dependency task completed → Auto-unblock
• Resource now available → Auto-retry
• Info supplemented → Auto-advance

For "Overdue Tasks"

Analysis Steps:

1. Assess remaining workload
2. Determine delay impact
3. Generate adjustment plan

Auto-adjust:

• Update deadline to reasonable value
• Adjust downstream task schedule
• Mark as high priority

Output Rules

Usage Examples

# Manual trigger detection
node skills/task-detection-thinking/scripts/detect.js

# View detection results
cat memory/hot/task-alert.md

# View thinking log
cat memory/hot/thinking-log.md

Configuration Parameters

task_detection:
  # Detection frequency
  heartbeat_check: true      # Detect during Heartbeat
  cron_schedule: "0 */6 * * *"  # Every 6 hours
  
  # Threshold settings
  stale_threshold_hours: 24   # Stall threshold
  overdue_check: true         # Check overdue
  dependency_check: true      # Check dependency chain
  
  # Auto-fix
  auto_fix_enabled: true      # Enable auto-fix
  auto_fix_max_attempts: 3    # Max auto-attempts
  
  # Notification settings
  feishu_alert: true          # Critical alert push
  alert_on_overdue: true      # Overdue alert
  alert_on_blocking: true     # Serious block alert

Workflow

Heartbeat trigger / Scheduled trigger / Manual trigger
    ↓
Scan HEARTBEAT.md + WORKING.md
    ↓
Identify anomaly tasks (stalled/blocked/overdue/duplicate)
    ↓
Analyze causes (context + historical memory)
    ↓
Generate solutions (3 options)
    ↓
Attempt auto-fix (no-human-needed solutions)
    ↓
Write detection results + thinking log
    ↓
Push critical alerts to Feishu

Integration with Existing Systems

• HEARTBEAT.md - Read global task board
• WORKING.md - Read subtask details
• Three-layer memory system - Store detection results and thinking logs
• ClawMemory - Query historical solutions
• Feishu - Push critical alerts

*Transform task management from reactive response to proactive prevention*