todokan-review-loop
由 banatealab 提供
Productivity & Tasks简介
Process Todokan task and thought boards with a review-loop workflow.
描述
该技能的功能
Monitors Todokan kanban boards for tasks in 'doing' status that users have explicitly sent to AI, reads the full comment thread, and posts a context-aware reply targeting the latest unresolved question. After commenting, it moves the task to 'done' for human review. Designed for recurring polling and cron-based automation via Todokan MCP.
The thread-aware reply logic tracks which questions have already been answered, so the agent targets the newest unresolved user comment rather than restating earlier answers.
何时使用
- Responding to AI-assigned Todokan tasks during off-hours
- Answering follow-up questions on in-progress board items without repeating prior answers
- Running scheduled review loops across multiple Todokan boards and habitats
- Posting summarized research results as structured task documents or comments
- Keeping multi-turn task threads accurate when a user corrects or changes their request
示例工作流
User asks: 'What CRM strategy should we use?' in a Todokan task, then adds a follow-up comment: 'How do I start tomorrow?'
- 1
Lists habitats and boards, finds tasks with status 'doing' and aiEnabled: true
- 2
Reads the full comment thread chronologically and maps each user question to whether it has been resolved by a subsequent MCP reply
- 3
Identifies 'How do I start tomorrow?' as the active unresolved question via turn-lock preflight
- 4
Posts a concise comment answering the concrete next-step question directly in the first line
- 5
Moves task status from 'doing' to 'done' for human review
A direct comment answering the follow-up question posted to the Todokan task thread, with task status set to 'done' (Review)
前置要求
- Todokan account at https://todokan.com
- TODOKAN_API_KEY environment variable
- TODOKAN_MCP_URL environment variable
安装
clawhub install banatealab/todokan-review-loop运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。