task-tracker（任务跟踪器）

Task Tracker - ADHD State Machine

Overview

Systematic task tracking designed for ADHD patterns. Prevents task abandonment through proactive interventions while maintaining accountability without judgment.

State Machine

                    ┌──────────────────────────────────────────┐
                    │                                          │
                    ▼                                          │
┌──────────┐   ┌──────────────────┐   ┌─────────────┐         │
│ INITIATED│──▶│ SOLUTION_PROVIDED│──▶│ IN_PROGRESS │─────────┤
└──────────┘   └──────────────────┘   └─────────────┘         │
     │                  │                    │                 │
     │                  │                    ▼                 │
     │                  │              ┌───────────┐          │
     │                  │              │ COMPLETED │          │
     │                  │              └───────────┘          │
     │                  │                                      │
     │                  ├─────────────────┐                   │
     │                  │                 │                    │
     │                  ▼                 ▼                    │
     │           ┌───────────┐     ┌──────────┐               │
     └──────────▶│ ABANDONED │     │ BLOCKED  │               │
                 └───────────┘     └──────────┘               │
                        │                │                     │
                        ▼                ▼                     │
                 ┌───────────┐     ┌──────────┐               │
                 │  DEFERRED │◀────│ (Retry)  │───────────────┘
                 └───────────┘     └──────────┘

State Definitions

Task Metadata

Every task tracks:

task = {
    "id": "task_20251219_001",
    "description": "Review Q4 insurance renewals",
    "state": "INITIATED",
    "domain": "BUSINESS",  # BUSINESS | MICHAEL | FAMILY | PERSONAL
    "complexity": 6,       # 1-10 scale
    "clarity": 8,          # 1-10 scale (how clear were instructions)
    "estimated_minutes": 45,
    "actual_minutes": None,
    "initiated_at": "2025-12-19T10:30:00Z",
    "solution_provided_at": None,
    "completed_at": None,
    "abandonment_count": 0,
    "intervention_level": 0
}

Abandonment Detection Triggers

Monitor for these patterns:

1. Context Switch - New topic without closing current task
2. Session End - Chat ends with task in SOLUTION_PROVIDED or IN_PROGRESS
3. Time Decay - >30 minutes since solution with no update
4. Topic Drift - User asks unrelated question while task pending

Detection Algorithm

def check_abandonment(task, current_message, elapsed_minutes):
    if task["state"] not in ["SOLUTION_PROVIDED", "IN_PROGRESS"]:
        return False

    # Time-based trigger
    if elapsed_minutes > 30:
        return True

    # Context switch detection
    if not is_related_to_task(current_message, task):
        return True

    return False

Intervention Protocol

Level 1: Gentle Check (0-30 minutes)

📌 Quick check: [task description] - still on it?

Level 2: Pattern Observation (30-60 minutes)

🔄 I notice [task] from earlier. This is pattern #[N] this week.
Continue or consciously defer?

Level 3: Direct Accountability (>60 minutes)

⚠️ ACCOUNTABILITY: [task] started [time] ago.
Current status? Be honest - no judgment, just facts.

ADHD Intervention Strategies

When abandonment detected, apply these techniques:

Micro-Commitment

Break to smallest possible action:

Just step 1? [tiny specific action] That's it.
Nothing else required right now.

Body Doubling

Create virtual co-working presence:

Let's do this together. You: [specific action]
Me: ⏱️ Waiting... (I'll check back in 5 minutes)

Chunking

Focus tunnel vision:

Step 1 only. Confirm when done.
Don't think about step 2 yet.

Energy Matching

Align with daily patterns:

It's [time] - your [peak/dip] period.
[Suggest appropriate task complexity]

Workflow: Task Creation

Trigger: Any request that implies work to be done

1. Extract task description from message
2. Assess complexity (1-10) based on:

- Number of steps - External dependencies - Decision-making required - Time estimate

1. Assess clarity (1-10) based on:

- Specificity of request - Known vs unknown elements - Ambiguity level

1. Assign domain (BUSINESS/MICHAEL/FAMILY/PERSONAL)
2. Set state to INITIATED
3. Provide solution → state to SOLUTION_PROVIDED

Workflow: State Transitions

INITIATED → SOLUTION_PROVIDED

# After providing solution
task["state"] = "SOLUTION_PROVIDED"
task["solution_provided_at"] = now()

SOLUTION_PROVIDED → IN_PROGRESS

Trigger: User says "working on it", "starting now", "doing this"

task["state"] = "IN_PROGRESS"

IN_PROGRESS → COMPLETED

Trigger: User says "done", "finished", "completed"

task["state"] = "COMPLETED"
task["completed_at"] = now()
task["actual_minutes"] = elapsed_time()
update_streak()

Any → ABANDONED

Trigger: Abandonment detection + no response to intervention

task["state"] = "ABANDONED"
task["abandonment_count"] += 1
log_pattern()

Any → BLOCKED

Trigger: User identifies external dependency

task["state"] = "BLOCKED"
task["blocker"] = identified_dependency
schedule_retry()

Any → DEFERRED

Trigger: Conscious postponement with reason

task["state"] = "DEFERRED"
task["defer_reason"] = reason
task["defer_until"] = scheduled_time

Streak Tracking

Maintain completion streak for motivation:

streak = {
    "current": 5,  # Days with at least 1 completion
    "longest": 12,
    "total_completions": 47,
    "abandonment_rate": 0.23  # 23% tasks abandoned
}

# On completion
"✅ Done. Streak: 5 days. Total: 47 tasks."

# On first task of day
"Day 6 starts now. Keep the streak alive."

Domain-Specific Behavior

BUSINESS (Everest Capital)

• Higher urgency weighting
• Deadline awareness
• Revenue impact assessment

MICHAEL (D1 Swimming)

• Connect to swim schedule
• Nutrition/training alignment
• Recruiting deadlines

FAMILY (Orthodox Observance)

• Shabbat/holiday awareness
• No work pressure evenings
• Family event priority

PERSONAL (Health/Learning)

• Energy level consideration
• Learning capture integration
• Health metric correlation

LangGraph Integration

State Schema

class TaskState(TypedDict):
    task_id: str
    description: str
    state: str
    domain: str
    complexity: int
    clarity: int
    elapsed_minutes: int
    intervention_level: int
    streak_current: int

Nodes

• create_task - Initialize from user message
• provide_solution - Generate response, update state
• check_abandonment - Periodic abandonment check
• intervene - Apply intervention strategy
• complete_task - Log completion, update streak

Scripts

• scripts/task_state_machine.py - State transition logic
• scripts/abandonment_detector.py - Pattern detection
• scripts/streak_calculator.py - Streak maintenance

References

• references/intervention_templates.md - Full intervention scripts
• references/adhd_patterns.md - Common patterns and responses