A to-do list is an ordered execution plan that externalizes goals, progress, and priorities. For AI agents, it acts as a context anchor: intermediate steps and previous errors can crowd out the original objective during long sessions. section 3.5.2
Use the list as an actively maintained control surface:
- Track task status as work progresses. section 3.5.2
- Check off, reprioritize, and update items rather than treating the plan as static. section 3.5.2
- Recite the plan into the freshest context at intervals, while recognizing that repeated reminders can eventually become limiting as models improve. section 3.5.2
A no-op with operational value
Claude Code’s to-do tool does not compile, lint, or execute code. It generates and maintains the execution plan, keeping high-level objectives visible without replaying the entire conversation history. section 3.5.2
Figure 3.9 Execution plan of Claude Code.

Inspect the execution-plan view to see how tasks and statuses provide a structured representation of progress. section 3.5.2
