Load Local Context#

bash

CONTEXT_FILE="/opt/ocana/openclaw/workspace/skills/pa-ownership/.context"
[ -f "$CONTEXT_FILE" ] && source "$CONTEXT_FILE"
# Then use: $OWNER_PHONE, $TASKS_FILE, $WORKSPACE, etc.

PA Ownership Skill#

Heleni tracks tasks she owns — executing, retrying when blocked, and closing the loop when done.

---

When to Use#

Trigger phrases:

• "take ownership"
• "track this" "own this"
• "add to my tasks"
• Any task Heleni explicitly commits to completing

---

Data File#

All tasks are persisted to:

text

/opt/ocana/openclaw/workspace/data/pa-tasks.json

Schema#

json

{
  "tasks": [
    {
      "id": "task_<YYYYMMDD_HHMMSS>",
      "title": "Short task description",
      "description": "Full context / what needs to happen",
      "status": "NEW",
      "initiated_by": "Netanel | group:<jid> | self",
      "created_at": "2026-04-03T10:00:00Z",
      "updated_at": "2026-04-03T10:00:00Z",
      "due_at": null,
      "attempts": 0,
      "max_attempts": 3,
      "last_attempt_at": null,
      "blocked_reason": null,
      "result": null,
      "reported_done": false
    }
  ]
}

Status Values#

StatusMeaningNEWTask received, not yet startedIN_PROGRESSCurrently executingDONECompleted successfullyBLOCKEDCannot proceed — waiting on somethingFAILEDExhausted all retry attempts

---

Step-by-Step Process#

Step 1: Register the Task#

When a task is received and Heleni commits to it:

1. React 👍 immediately to the owner's message (before starting work)

2. Read data/pa-tasks.json (create if missing: {"tasks": []})

3. Generate a task ID: task_<YYYYMMDD_HHMMSS>

4. Determine initiated_by:

   • DM from Netanel → "Netanel"
   • Group message → "group:<jid>"
   • Heleni's own initiative → "self"

5. Set status: "NEW", attempts: 0

6. Write updated JSON back to file

7. Write to WhatsApp memory: memory/whatsapp/dms/<PHONE-sanitized>/context.md

Step 2: Execute#

1. Set status to IN_PROGRESS, update updated_at
2. Execute the task using available tools
3. On success → go to Step 4 (Done)
4. On failure → go to Step 3 (Retry)

Step 3: Retry (Blocked / Failed Attempt)#

On failure or block:

1. Increment attempts counter
2. Record blocked_reason
3. Record last_attempt_at

Backoff schedule:

• Attempt 1 fail → retry after ~5 minutes
• Attempt 2 fail → retry after ~15 minutes
• Attempt 3 fail → mark as FAILED, notify Netanel immediately

text

If attempts >= max_attempts:
  → Set status: "FAILED"
  → Report to initiated_by: "❌ [task] failed after 3 attempts: [reason]"
  → Do NOT retry further
Else:
  → Set status: "BLOCKED"
  → Log blocked_reason
  → Schedule retry (via heartbeat or mental note)

Step 4: Mark Done#

1. Set status: "DONE", result: "<outcome summary>", updated_at: now

2. If reported_done == false:

   • React ✅ to the original task message
   • Report to initiated_by with result
   • Set reported_done: true

3. Update the JSON file

4. Write outcome to WhatsApp memory file

Close-the-loop rule: ALWAYS report back to whoever initiated the task. No exceptions.

• Netanel initiated → send DM to Netanel
• Group initiated → reply in that group
• Self-initiated → log in daily memory

Step 5: Write to File#

After every state change, write the full updated pa-tasks.json.

python

import json, datetime, os

TASKS_FILE = "/opt/ocana/openclaw/workspace/data/pa-tasks.json"

def load_tasks():
    if not os.path.exists(TASKS_FILE):
        return {"tasks": []}
    with open(TASKS_FILE) as f:
        return json.load(f)

def save_tasks(data):
    os.makedirs(os.path.dirname(TASKS_FILE), exist_ok=True)
    with open(TASKS_FILE, "w") as f:
        json.dump(data, f, indent=2, ensure_ascii=False)

def update_task_status(task_id, status, **kwargs):
    data = load_tasks()
    for task in data["tasks"]:
        if task["id"] == task_id:
            task["status"] = status
            task["updated_at"] = datetime.datetime.utcnow().isoformat() + "Z"
            for k, v in kwargs.items():
                task[k] = v
            break
    save_tasks(data)

---

Heartbeat Integration#

During every heartbeat, scan for stale tasks:

text

For each task in pa-tasks.json where status IN ["IN_PROGRESS", "BLOCKED"]:
  age = now - updated_at

  If age > 2 hours AND NOT reported_stuck:
    → Mark last_notified_at = now

  If status == "BLOCKED" AND attempts < max_attempts:
    → Attempt retry

Add this block to HEARTBEAT.md:

text

## PA Ownership Check
- Scan data/pa-tasks.json for IN_PROGRESS or BLOCKED tasks
- Alert if any task >2h without update
- Retry BLOCKED tasks with remaining attempts

---

HEARTBEAT.md Snippet#

When setting up this skill for the first time, add to HEARTBEAT.md:

markdown

## Task Ownership Check
- Read /opt/ocana/openclaw/workspace/data/pa-tasks.json
- Flag any task with status IN_PROGRESS or BLOCKED updated >2h ago
- Retry BLOCKED tasks if attempts < max_attempts
- Report FAILED tasks to Netanel immediately

---

Alert Format#

Task Stuck Alert (>2h)#

text

📋 [task title]
[max_attempts]

Task Complete#

text

📝 [result summary]

Task Failed (all retries exhausted)#

text

---

Rules#

1. Always close the loop — when done, always report to whoever initiated it
2. Never silently fail — BLOCKED or FAILED always triggers a notification
3. Max 3 retries — after that, escalate to Netanel
4. Persist everything — every state change is written to pa-tasks.json
5. Heartbeat integration — stale tasks surface automatically, no manual polling

---

Cost Notes#

• File reads/writes: cheap — do on every state change
• Retry logic: don't call external APIs in tight loops; space out via heartbeat
• Alert: only send once per "stuck" detection (use last_notified_at)