To-Do Accelerator
Collaborative task management between users and AI Agents. Agents pick up tasks, work on them, deliver results, and update progress — functioning as teammates with the user. The Obsidian Kanban board serves as the shared communication channel.
Setup
Check: If todo-accelerator-config.yaml exists in the agent's workspace → ready to use. If not → follow initialization.md in this directory (one-time process).
Working Scenarios
| Scenario | Agent Action |
|---|
| User describes a new task or bookmarks a topic | Ask for missing details (targets, requirements, priority) → INLINECODE2 |
| Agent heartbeat event |
Call
work-on-todo → process the returned prompt →
commit when done or blocked |
| User asks to handle a specific task | Call
list-pending → confirm with user →
work-on-todo --name "..." →
commit |
Commands
Base command (referred to as <CMD> below):
CODEBLOCK0
Replace <skill-dir> with the absolute path to this skill's directory, and <workspace> with the agent's workspace directory.
add-todo
Create a new to-do with a companion note, added under "Ideas".
CODEBLOCK1
Example:
CODEBLOCK2
Do not guess parameters. If the user hasn't specified targets or requirements, ask them:
- - "What outcome do you expect from this?"
- "Are there specific sub-tasks or questions to address?"
- "Should this be prioritized? (0 = normal, higher = more urgent)"
work-on-todo
Pick up the highest-priority to-do from Ideas and prepare it for processing.
CODEBLOCK3
Example:
CODEBLOCK4
Behavior:
- - Without
--name: auto-selects by priority (highest first, random among ties) - With
--name: works on that specific to-do (confirm exact name with user first) - If the selected to-do has no unchecked requirements → moves it to 审阅中 and skips (no action needed)
- Otherwise → moves card from Ideas to 推进中, increments iteration count, and returns a structured prompt
After work-on-todo returns, follow the instructions in references/processing-work-on-todo.md.
commit
Check off completed requirements and finalize the current round of work.
CODEBLOCK5
Example:
CODEBLOCK6
Each --completed string must exactly match an unchecked requirement in the note's "What's More" section. If any string doesn't match, the script returns an error with the remaining unchecked requirements — re-examine and pass the exact strings.
The script moves the card to 审阅中 after committing.
list-pending
List all to-dos under "Ideas" with their priority levels.
CODEBLOCK7
Example:
CODEBLOCK8
Note Structure
Each to-do has a companion .md note. For YAML frontmatter properties, see references/note-yaml-properties.md.
| Section | Purpose | How to update |
|---|
| What's More | Requirements checklist (- [ ] / - [x]) | Managed via commit command |
| Target |
Final results and deliverables | Write directly in the note |
|
Investigation and Problems | Ongoing findings, progress notes, obstacles | Write directly in the note |
Case Studies
See references/case-study.md for detailed usage examples covering the full lifecycle.
待办加速器
用户与AI智能体之间的协作任务管理。智能体可领取任务、执行任务、交付结果并更新进度——作为用户的团队成员协同工作。Obsidian看板作为共享沟通渠道。
设置
检查： 若智能体工作区中存在todo-accelerator-config.yaml → 即可使用。若不存在 → 按照本目录中的initialization.md操作（一次性流程）。
工作场景
| 场景 | 智能体操作 |
|---|
| 用户描述新任务或收藏某个主题 | 询问缺失细节（目标、要求、优先级）→ add-todo |
| 智能体心跳事件 |
调用work-on-todo → 处理返回的提示 → 完成或受阻时执行commit |
| 用户要求处理特定任务 | 调用list-pending → 与用户确认 → work-on-todo --name ... → commit |
命令
基础命令（下文简称）：
python3 <技能目录>/scripts/todo.py --config <工作区>/todo-accelerator-config.yaml
将<技能目录>替换为本技能目录的绝对路径，将<工作区>替换为智能体的工作区目录。
add-todo
创建带有配套笔记的新待办事项，添加至想法栏。
bash
add-todo --name 标题 [--targets 成果1 成果2] [--requirements 要求1 要求2] [--priority N] [--allow-subagent | --no-allow-subagent] [--assigned-agent 智能体ID]
示例：
bash
add-todo --name 研究AutoResearch --targets 功能摘要和设置指南 --requirements 阅读README 尝试运行演示 撰写发现报告 --priority 1
请勿猜测参数。 若用户未指定目标或要求，请询问他们：
- - 您期望达成什么成果？
- 是否有需要处理的特定子任务或问题？
- 是否需要设定优先级？（0 = 普通，数值越高越紧急）
work-on-todo
从想法栏中选取优先级最高的待办事项并准备处理。
bash
work-on-todo
work-on-todo --name 特定待办事项
示例：
bash
work-on-todo --name 研究AutoResearch
行为说明：
- - 不使用--name：按优先级自动选择（最高优先，同级随机）
- 使用--name：处理该特定待办事项（先与用户确认准确名称）
- 若所选待办事项无未勾选要求 → 移至审阅中并跳过（无需操作）
- 否则 → 将卡片从想法移至推进中，增加迭代次数，并返回结构化提示
work-on-todo返回后，请遵循references/processing-work-on-todo.md中的说明。
commit
勾选已完成的要求并结束当前轮次工作。
bash
commit --name 待办名称 --completed 要求1 要求2
示例：
bash
commit --name 研究AutoResearch --completed 阅读README 尝试运行演示
每个--completed字符串必须完全匹配笔记更多内容部分中未勾选的要求。若任何字符串不匹配，脚本将返回错误并显示剩余未勾选要求——请重新检查并传入准确的字符串。
提交后脚本会将卡片移至审阅中。
list-pending
列出想法栏中所有待办事项及其优先级。
bash
list-pending
示例：
bash
list-pending
笔记结构
每个待办事项都有一个配套的.md笔记。关于YAML前置属性，请参见references/note-yaml-properties.md。
| 章节 | 用途 | 更新方式 |
|---|
| 更多内容 | 要求清单（- [ ] / - [x]） | 通过commit命令管理 |
| 目标 |
最终成果和交付物 | 直接在笔记中编写 |
|
调查与问题 | 持续发现、进度记录、障碍 | 直接在笔记中编写 |
案例研究
完整生命周期使用示例请参见references/case-study.md。
