Your AI agent skips steps. Give it a todo list.
说明与用途
TodoWrite 工具通过将计划记录为检查表并确保一次只执行一个任务,防止代理跳过步骤,从而实现完整且可视化的任务执行。
问题
AI agents often skip steps in long tasks, leaving some tasks unfinished without errors.
结果
The TodoWrite tool prevents agents from skipping steps by maintaining a live checklist, ensuring only one task is in progress at a time, and updating the plan after each step. This leads to complete task execution and clear visibility of progress.
工作流
- Give the agent a multi-step request with a clear plan.
- The agent calls TodoWrite to generate a todo list with tasks and statuses.
- The agent processes one task at a time, marking it in_progress.
- After completing a task, the agent calls TodoWrite again to update the list, marking the task completed.
- The agent repeats until all tasks are completed.
- The agent returns a final summary of the work done.
证据
Language models pay less attention to whatever sits in the middle of a long context. As an agent works, its context fills with tool calls and results, and the original plan drifts into that middle zone.
One rule does most of the work: only one task can be in_progress at a time.
If the agent skips the list on a big job, make the request more explicit about its steps or tighten the system prompt.
- 记录日期: 2026-11-02T00:43:51.999Z