# Todo Write

> TodoWrite-driven plan that the agent commits to before generation.

- Skill: `nexu-io/todo-write` (Agent Skill)
- Install (CLI): `npx skillmds@latest add nexu-io/todo-write`
- Raw SKILL.md: https://api.skillmd.com/api/skills/nexu-io/todo-write/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: Nexu (https://skillmd.com/u/nexu-io)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/nexu-io/todo-write

---


# Todo write

Before writing any artifact files, the agent commits to a numbered plan
via its plan tool. The plan is the audit trail; subsequent turns
either tick items off or rewrite the plan. The atom's prompt fragment
teaches the agent to:

1. Keep todos atomic (one verb per todo).
2. Reorder freely as the picture sharpens.
3. Mark a todo complete only after the matching artifact lands.
4. Surface blockers as todos — never silently skip.

The OpenDesign daemon does not enforce a particular tool name; use
whichever plan tool this session actually exposes. The host renders
every one of them as the same live Todos card.
The atom's job is to keep "make a plan first" in the system prompt so
non-trivial workflows don't skip the planning step.