# Ask User

> Intercept every agent-originated question and unresolved ambiguity that requires user input. ALWAYS use before asking for clarification, confirmation, permission, preference, approval, waiver, a decision, or any answer that work must wait for. Do not use when no user response is required; answer user-originated factual questions directly, ignore rhetorical questions, and settle discoverable or reversible matters yourself.

- Skill: `jcosta33/ask-user` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jcosta33/ask-user`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jcosta33/ask-user/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: jcosta33 (https://skillmd.com/u/jcosta33)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jcosta33/ask-user

---


# Ask User

If the next step needs a user response, stop.

1. Exhaust discoverable facts and reversible conventions.
2. State one blocked choice.
3. Use the native picker: recommendation first, three genuine options by default, two when binary,
   one plain reason and cost each, and `Other`.
4. Without a picker, render the same numbered choices plus `Other`.
5. Batch independent choices. Resolve dependent choices in order. Freeze dependent work.

Never infer preference. Never ask bare prose. Never disguise a question as "let me know," "should
I," or "would you like me to."