AgentSkillsCN

task-engine

从简报或脑力激荡中创建、组织与优先排序任务。与Taskmaster MCP同步。用于规划工作、跟踪状态,以及挑选下一步行动。

SKILL.md
--- frontmatter
name: task-engine
description: Create, organize, and prioritize tasks from briefs or brain dumps. Sync with Taskmaster MCP. Use to plan work, track status, and pick next steps.
triggers:
  - create tasks
  - organize tasks
  - what needs to be done
  - prioritize
  - תארגן את המשימות
  - brain dump

Task Engine Skill

Transform project briefs and brain dumps into organized, prioritized task lists with dependencies. Syncs bidirectionally with Taskmaster MCP.

Purpose

The task engine manages all task-related operations including creation, organization, prioritization, and synchronization with Taskmaster MCP. It supports both structured input (from briefs) and unstructured input (brain dumps).

Taskmaster MCP Integration

Read Operations

FunctionPurpose
get_tasksRetrieve all tasks for a project
get_taskGet specific task by ID
next_taskFind next task to work on based on dependencies

Write Operations

FunctionPurpose
parse_prdGenerate tasks from PRD document
expand_taskBreak task into subtasks
set_task_statusUpdate task status (pending/in-progress/done)
update_subtaskModify subtask details

Sync Direction

  • Local → Taskmaster: Push changes from {project}_tasks.md to MCP
  • Taskmaster → Local: Pull updates from MCP to {project}_tasks.md
  • Bidirectional: Keep both in sync

Inputs

Required:

  • source: Brief, brain dump, or raw task list

Optional:

  • project_context: From {project}_brief.md
  • existing_tasks: From {project}_tasks.md or Taskmaster

Workflow

Step 1: Capture Tasks

From Brief:

  • Extract implied tasks from milestones
  • Convert deliverables to tasks
  • Identify dependencies from success criteria

From Brain Dump:

  • Parse text for action items
  • Identify verbs indicating tasks
  • Group related items

From Existing List:

  • Clean up format
  • Standardize structure
  • Fill missing fields

Step 2: Structure Tasks

Apply these rules:

  • One clear outcome per task
  • Create subtasks if work exceeds 2 hours
  • Maximum 3 levels of nesting
  • Each task needs: ID, name, status, priority

Step 3: Analyze Priority

Use the Eisenhower Matrix:

QuadrantPriorityAction
Urgent + ImportantHigh PriorityDo first
Not Urgent + ImportantMedium PrioritySchedule
Urgent + Not ImportantDelegateAssign out
NeitherLow Priority / BacklogDo later

Step 4: Identify Dependencies

For each task, determine:

  • What must complete before this?
  • What does this block?
  • Are there parallel paths?

Step 5: Sync with Taskmaster

python
# Example sync workflow
tasks = taskmaster.get_tasks(project_root)
local_tasks = parse_tasks_md(f"{project}_tasks.md")

# Merge and update
for task in merged_tasks:
    if task.changed_locally:
        taskmaster.set_task_status(task.id, task.status)
    elif task.changed_remotely:
        update_local_task(task)

Step 6: Checkpoint

CHECKPOINT: Present organized list to user

Ask:

  • "Does this capture everything?"
  • "Is the priority correct?"
  • "Any dependencies I missed?"

Step 7: Save

  • Save as {project}_tasks.md
  • Suggest: "Use delegation-advisor to assign high priority tasks"

Output Format

See resources/tasks_template.md for full template.

markdown
# Tasks: {project_name}

**Last Updated:** {timestamp}
**Total Tasks:** {count} | **Completed:** {done_count}

## High Priority / Urgent

### [ ] {task_name}
- **ID:** {task_id}
- **Status:** {status}
- **Assignee:** {human|ai_agent|pending}
- **Dependencies:** {deps}

## Medium Priority
...

## Low Priority / Backlog
...

## Completed
- [x] {completed_task} (Done: {date})

Example: Brain Dump Processing

User Input:

"brain dump: צריך לחקור frameworks קיימים, לכתוב draft לדיקן, להכין מצגת, לתאם פגישה עם IT, לבדוק תקציב"

Processing:

  1. Extract 5 action items
  2. Categorize by type (research, writing, meetings)
  3. Assign priorities based on dependencies
  4. Sync with Taskmaster

Output:

markdown
# Tasks: AI Strategy

## High Priority / Urgent
### [ ] Research existing AI frameworks
- **ID:** task-1
- **Status:** pending
- **Dependencies:** none

### [ ] Draft proposal for Dean
- **ID:** task-2
- **Dependencies:** task-1

## Medium Priority
### [ ] Prepare presentation
- **ID:** task-3
- **Dependencies:** task-2

Cross-Interface Notes

  • Claude AI: Taskmaster MCP called via skill
  • Claude Code: Taskmaster MCP called natively
  • Both use same task format for compatibility