todero-ask-or-decide

Limits clarifying questions to three and decides remaining assumptions autonomously during planning.

Updated Mar 26, 2026
One-click install
npx skills add https://github.com/nabitllc/todero --skill todero-ask-or-decide-nabitllc
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: todero-ask-or-decide
Source: https://github.com/nabitllc/todero/tree/main/skills/todero-ask-or-decide
Command: npx skills add https://github.com/nabitllc/todero --skill todero-ask-or-decide-nabitllc

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents often stall work by asking endless clarifying questions one at a time, stranding the person who must answer them. This Skill enforces a discipline of asking at most three questions, once, at the start, and deciding everything else with stated assumptions. ## Core Features & Use Cases - Question filtering: Every candidate question must pass the test of whether its answer would change the plan; otherwise the agent decides and records the assumption in one line. - Recommended answers: Each question is paired with the agent's own recommended answer so the person can confirm quickly instead of deliberating. - Confidence discipline: Facts are stated only at 95% confidence or higher, vague replies like "sounds good" are not treated as agreement, and jargon-heavy questions are decided rather than asked. - Use Case: During planning or drafting tasks, an agent receives a vague goal, asks the three questions that matter (scope, definition of done, hard limits), then proceeds autonomously with documented assumptions. ## Quick Start Apply the ask-or-decide rules to this task: ask me at most three questions whose answers would change the plan, then proceed with stated assumptions.

Frequently Asked Questions about todero-ask-or-decide

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stop an AI agent from asking too many questions?

Limit the agent to at most three questions asked once at the start, and require each question to pass a test: would the answer change the plan? Everything else should be decided by the agent with the assumption stated in one line.

How should an AI agent handle vague answers like sounds good?

A vague reply is not agreement. The agent should state which answer it is taking and why in one line, then continue working rather than treating the ambiguity as resolved.

What questions should an agent ask before starting a task?

The three that usually matter are scope (what must be in the first version), done (what completion looks like), and limits (hard constraints on money, time, or people). Each question should include the agent's own recommended answer.

When should an agent decide instead of asking the user?

Decide when the answer would not change the plan, or when the question cannot be phrased without jargon the person would understand. The agent reports the decision in one sentence instead of stalling.

How confident should an agent be before stating something as fact?

State something as fact only at 95 percent confidence or higher. Below that threshold, the agent should say it is unsure and name what it is unsure about rather than bluffing.