multiplechoice
Skill1 starsChanged 4 months agoMITView on GitHub
Use when the user asks for the decisions they need to make or confirm, what was decided on their behalf, or for options with your recommendation. Also use on your own when two or more open decisions or judgment calls block the work, or before long autonomous execution that depends on the user's choices.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- biascheckSkill
- crew-consultSkill
- crew-reviewSkill
- experimentSkill
- priorartSkill
- secure-researchSkill
- unbiasSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
Every open and made-for-you item becomes a picker question.
- Distinctive: line 32
Verify the facts each option depends on (grep, file reads, a quick command).
- Distinctive: line 34
When a question rests on something the user said, quote their words and keep their scope.
- Distinctive: line 36
Keep each option to a single premise.
- Distinctive: line 64
When every round is done, give a short table (decision → choice) and record it wherever the project keeps decisions.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
multiplechoice
Overview
Turn every open decision and every consequential call you made on the user's behalf into multiple-choice questions, with your recommendation first, asked in rounds the user can answer. The user can only decide what reaches a picker. A call you describe in prose never gets a vote.
1. Build the decision list
Collect all of the following, then sort them:
- Open: a choice you can't make from the request, the code, or a sensible default.
- Made for you: anything you decided without asking that a reasonable user might reverse. That includes tool or library picks, defaults, scope cuts, deletions, and readings of an ambiguous instruction.
- Test-decidable: a choice that measurement or an experiment you can run would settle. Don't ask the user to guess these. Propose the experiment as one question (run it, or skip it and take the default), and ask the resulting choice later, with the numbers.
Every open and made-for-you item becomes a picker question. Don't leave any of them as prose "defaults unless you object". A clear correctness bug is not a decision: fix it and state it in one line.
2. Check before you ask
- Verify the facts each option depends on (grep, file reads, a quick command). Don't frame options from memory.
- When a question rests on something the user said, quote their words and keep their scope. Don't widen or paraphrase it.
- Keep each option to a single premise. Split bundled actions such as "merge and tag" into separate options or questions.
3. Ask, one round per turn
Each question has:
- stem: the decision, plus the facts it turns on (numbers, file:line, cost), in plain words. Mark calls you made with "(made for you)".
- header: a chip of 12 characters or fewer.
- 2–4 options: the recommended option first, with "(Recommended)" in its label. Each option's description gives a one-line consequence.
Send at most 4 questions per round, most blocking first. Open with one line of context, such as "Round 1 of 3: the four that unblock the merge". Send one round, then wait for the answers before the next.
Without a picker tool (such as AskUserQuestion), print the same round as a numbered list with
lettered options and your pick marked, then ask the user to reply like 1a 2c 3 other: ….
4. After the answers
- "Other" with a twist or a new option: adopt it as stated.
- A question back ("what does X use?"): answer it, then re-ask that decision with the new facts in the stem.
- A rejected premise: fix the premise and re-ask. Don't argue for the original.
- A non-recommended pick: take it without re-arguing.
When every round is done, give a short table (decision → choice) and record it wherever the project keeps decisions. Flag any choice that reverses an earlier decision, then continue the work.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

