aif-grounded

Require 100/100 confidence and evidence before answering high-stakes questions.

Updated Mar 23, 2024
One-click install
npx skills add https://github.com/Ard2p/sk-bar-site --skill aif-grounded-ard2p
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: aif-grounded
Source: https://github.com/Ard2p/sk-bar-site/tree/main/.cursor/skills/aif-grounded
Command: npx skills add https://github.com/Ard2p/sk-bar-site --skill aif-grounded-ard2p

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Reliability gate that prevents guessing by requiring explicit evidence and 100/100 confidence before presenting conclusions.

Core Features & Use Cases

  • Enforces evidence-based reasoning and explicit uncertainty for high-stakes questions.
  • Outputs a concise “what’s missing” checklist when confidence is below 100.
  • Supports project-specific skill-context rules to tailor behavior and guarantees.

Quick Start

Provide an answer only after achieving 100/100 confidence based on available evidence.

Frequently Asked Questions about aif-grounded

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

FAQPage Schema
How do I prevent AI from guessing on high-stakes security and legal questions?

The reliability gate enforces explicit uncertainty by outputting a concise what's missing checklist when confidence is below 100, detailing the specific evidence needed to reach a grounded conclusion.

What is an evidence-based reliability gate for high-stakes domains?

It applies to contexts where facts may change or require verification, such as medical, security, legal, or policy questions, ensuring outputs are grounded rather than guessed.

Can I use project-specific rules to tailor no-guessing behavior for medical or policy questions?

This ensures the enforced reasoning and explicit uncertainty requirements align with your specific project constraints and verification needs.

What's the best way to enforce 100/100 confidence before an AI answers?

This approach ensures explicit uncertainty is maintained and prevents the AI from presenting conclusions without complete evidence-based verification.

How do I get a checklist of missing evidence when AI confidence is low?

This checklist details the exact missing information required to achieve 100/100 confidence and produce a grounded, evidence-based conclusion.

When should I not use an evidence-based reliability gate for AI responses?

Applying this no-guessing constraint to low-stakes contexts may unnecessarily block answers by demanding explicit evidence where uncertainty is acceptable.