Guide · Agent skill
How to use the Ask User Choice Gate agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The ask-user skill is a reusable pattern that teaches an AI agent to stop and ask you to choose between a few explicit options when a decision matters, instead of guessing and continuing. It comes from the GBrain project, where other skills reuse it at decision points such as filing notes or deleting files. After this guide you will know what the skill changes in the agent's behaviour, how to install it, how to phrase requests that trigger it, and what to watch for.
Ask User Choice Gategarrytan/gbrain · MIT Download the skill Quick start
- 1Download the ZIP. Get the ask-user ZIP from this skill's page on the directory.
- 2Unzip into your skills folder. Extract it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Ask for a choice gate. Use a request such as "present options" or "ask before proceeding" in your message.
- 5Answer with a click or short reply. Reply with a button, a number, or a phrase like "the first one"; the agent acknowledges and branches.
- 6Decline if needed. Choose Skip or Cancel, or say none of these, to stop that step without any action.
What the ask-user skill makes the agent do
Without this skill, an agent facing an ambiguous or risky choice may pick a default and carry on. The ask-user skill sets a consistent contract: the agent presents 2-4 labeled options, always includes an escape hatch such as Skip or Cancel, and then stops its turn. Nothing else runs until you reply.
The skill also explains how to read your answer. A button click, a number, or a typed phrase like "the first one" is mapped to the matching option, and the agent acknowledges briefly before acting on it. If your message is unrelated, the skill says the gate is dropped.
- Presents 2-4 options with self-explanatory labels such as "Merge — combine with existing page"
- Always offers an escape hatch (Skip, Cancel, or "none of these")
- Stops the turn after the question and makes no follow-up tool calls
- Asks one question per message and never stacks choice gates
Install the ask-user skill
You can install the skill in either of two ways. Download the ZIP from the skill's page on this site and unzip it into your personal skills folder at ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it limited to one project. The ZIP contains a single file, SKILL.md.
Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload
~/.claude/skills/ask-user/SKILL.mdTrigger the skill with natural requests
The skill's metadata lists trigger phrases. Asking for these in plain language tells the agent to use the pattern: "present options", "ask before proceeding", "choice gate", or "user decision". You can also ask for it indirectly by describing a decision you want to make yourself.
Other GBrain skills call this pattern on their own at decision points, so you may see it appear during tasks like importing contacts or filing meeting notes without asking for it.
- "Present options for how to file these meeting notes before you do anything."
- "Ask before proceeding with the bulk delete, and give me a Skip option."
- "Use a choice gate before you overwrite the existing page."
Examples of what the agent shows you
For a destructive operation, the skill's example format leads with a warning emoji, a one-line summary, and labeled choices such as "Delete them", "Archive first", "Keep them", and "Show me the list". For a filing decision, it leads with a routing emoji and offers options such as merging into an existing person's page or creating a new meeting page.
The context block is kept to 1-3 lines so you can understand the decision quickly. Your reply, whether a click or a short typed answer, is the only thing that lets the agent continue.
What the skill needs and how it behaves on other platforms
The skill lists no requirements, does not run scripts, and does not use the network. It is written to work anywhere an agent can send a message: the plain numbered format works in any chat or CLI, while interactive buttons are described for platforms such as Telegram, Discord, and Slack. Some OpenClaw agents have a built-in clarify tool that presents the same choices natively.
When to skip the gate: the skill tells the agent not to ask for clear instructions, low-stakes decisions, time-critical operations, or when you have already stated a preference.
- Requirements: none listed
- Runs scripts: no
- Uses network: no
- Works with plain text, inline buttons, or a clarify tool where available
Safety notes for gated actions
The skill is written so that a choice you have not made does not happen. The agent must not pick a default and proceed, and it should not make assumptions about your answer. If a gbrain call returns a confirmation-required result, the skill says to relay the prompt as the choice and wait. It also says the agent must never pass confirmation flags such as --yes or --max-usd on your behalf to move on. If you do not answer, the step is reported as skipped.
- The agent waits for you; silent defaults are treated as a failure of the pattern
- Confirmation flags are not supplied on your behalf
- Skipped steps are reported rather than quietly completed
What people use it for
- Confirming which page to file meeting notes under before anything is written
- Pausing before a bulk delete or overwrite and offering Archive first or Keep them
- Triage: letting you pick which task the agent should work on first
- Developers building their own GBrain-style skills who want a standard input step
Frequently asked questions
Does the ask-user skill cost money?
The source does not state a price or licence terms beyond the listing's MIT licence, so check the project's repository for details. The skill itself needs no requirements and runs no scripts.
What do I need to run the ask-user skill?
The skill lists no requirements. You need an agent that loads skills from ~/.claude/skills/ or .claude/skills/, or the Claude apps with Capabilities → Skills available.
Where do I put the ask-user files?
Unzip the downloaded ZIP so the skill folder sits in ~/.claude/skills/ for all projects, or in a project's .claude/skills/ folder for one project only. The package includes SKILL.md.
Can I use ask-user with Telegram or Discord?
Yes. The skill describes inline buttons for platforms such as Telegram, Discord, and Slack, and falls back to a numbered list on any platform.
Why won't my agent ask me anything?
Check that your request is ambiguous or risky enough. The skill tells the agent not to gate on clear instructions, low-stakes choices, or time-critical work, and to skip it when you have stated a preference.
Is ask-user a replacement for other agent skills?
No. It is a shared pattern that other skills, such as cold-start, ingest, and enrich in the GBrain project, call at decision points. Use it when you want a consistent way to ask for input rather than inventing one.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Choose the ask-user skill if you want your agent to confirm before destructive, ambiguous, or routing decisions, or if you are building GBrain-style skills that need a standard input step. Pick something else if you want fully autonomous runs, or if your tasks are simple and clear enough that the agent should just act.
Download the skill