1. Parse project, feature
   ↓
2. Find all tasks (in-progress first, then todo, sorted)
   ↓
3. Load SDD context
   ↓
4. Clarify ambiguities BEFORE execution
   ↓
5. Create feature branch (optional)
   ↓
6. Execute each task via faion-execute-task skill
   ↓
7. Run post-execution review (faion-tasks-reviewer-agent)
   ↓
8. Quality checks (tests, lint)
   ↓
9. Report summary

# Priority order
1. tasks/in-progress/TASK_*.md  # Resume first
2. tasks/todo/TASK_*.md         # Then new tasks

# Sort by task number
TASK_001, TASK_002, TASK_003...

for task in sorted_tasks:
    result = Skill(
        skill="faion-execute-task",
        args=f"{project}/{feature} {task.name}"
    )

    if result.status == "BLOCKED":
        if is_critical(result.error):
            STOP  # Git conflict, build broken, security
        else:
            CONTINUE  # Test failures, minor issues

    log_result(result)

Task(
    subagent_type="faion-tasks-reviewer-agent",
    prompt=f"Review completed tasks for {project}/{feature}"
)

## Feature Execution: {project}/{feature}

### Summary
- **Tasks:** {completed}/{total}
- **Status:** ✅ Complete | ⚠️ Partial | ❌ Failed
- **Duration:** {time}

### Task Results
| Task | Status | Commit | Notes |
|------|--------|--------|-------|
| TASK_001 | ✅ | abc123 | |
| TASK_002 | ✅ | def456 | |
| TASK_003 | ⚠️ | - | Tests failing |

### Quality Report
- Tests: {pass}/{total}
- Lint: {status}

### Next Steps
{recommendations}