you-sure

Extract factual claims from conversations and verify uncertain ones via adversarial debate.

4|1|Updated May 31, 2015

Search Tags:#verification #calibration #confidence #claims #ai-safety #adversarial-debate #calibration-report

Ready to run

npx skills add https://github.com/samcdavid/dotfiles --skill you-sure-samcdavid

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: you-sure
Source: https://github.com/samcdavid/dotfiles/tree/main/claude/skills/you-sure
Command: npx skills add https://github.com/samcdavid/dotfiles --skill you-sure-samcdavid

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Adversarial confidence challenge workflow that ensures claims are properly rated and independently verified before acting on recommendations, keeping conversations honest.

Core Features & Use Cases

Extract and catalog discrete factual claims from chat transcripts or conversations.
Rate initial confidence and trigger an adversarial-debate verifier for claims that are not Certain.
Generate a structured calibration report that summarizes verifications and recommends actions to ensure reliability.

Quick Start

Prompt the assistant to begin the claim extraction and verification workflow on the current conversation.

Frequently Asked Questions about you-sure

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

FAQPage Schema

How do I verify factual claims and calibrate confidence in an AI conversation?▼

Claim verification works by extracting every discrete factual claim from a conversation, assigning a self-rated confidence score, and launching an adversarial-debate agent to challenge any claim rated below Certain before returning a verdict.

What is an adversarial-debate agent for AI safety and claim verification?▼

An adversarial-debate agent is a verification mechanism that re-reads source material to challenge factual claims. It independently tests claims rated below Certain and returns a verdict to ensure conversation reliability before action is taken.

How do I generate a calibration report for verified conversation claims?▼

You generate a calibration report by compiling initial self-rated confidence scores, verification outcomes from the adversarial-debate process, and recommended actions into a structured summary that details the reliability of each claim.

Can I use this claim verification workflow on existing chat transcripts?▼

Yes, you can use this workflow on existing chat transcripts. The extraction process catalogs discrete factual claims from the conversation text, rates their initial confidence, and verifies them against source material without requiring real-time input.

What is the best way to ensure AI recommendations are reliable before acting on them?▼

The best way to ensure reliability is to run an adversarial confidence challenge that independently verifies claims through debate before acting on recommendations, producing a calibration report with explicit recommended actions.