judgment-guard

Calibrate AI advice confidence to match evidence quality and recency.

3|Updated Mar 31, 2026

Search Tags:#confidence calibration #assumption mapping #evidence grading #decision audit #risk-aware recommendations #fact vs inference

Ready to run

npx skills add https://github.com/fydel-ai/judgment-guard --skill judgment-guard

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: judgment-guard
Source: https://github.com/fydel-ai/judgment-guard/tree/main
Command: npx skills add https://github.com/fydel-ai/judgment-guard --skill judgment-guard

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Judgment-guard prevents AI advice from sounding more certain than the supporting evidence justifies, reducing overconfidence on high-stakes decisions.

Core Features & Use Cases

Evidence-to-confidence calibration: Assesses evidence strength and matches the answer’s expressed certainty to what the evidence supports.
Hidden assumption surfacing: Identifies unspoken premises and makes them explicit so users can evaluate the reasoning.
Decision-audit style output: Produces a structured report including a decision audit, claim map, calibrated answer, and verification checklist.
Intervention policy by evidence quality: Uses different rewriting strategies for strong, mixed, or weak evidence scenarios.
Use cases: Recommendations, rankings, comparisons, and prioritization on consequential topics such as money, health, career, legal, safety, and purchases.

Quick Start

Use judgment-guard to rewrite a draft recommendation with confidence matched to evidence for your consequential decision prompt.

Frequently Asked Questions about judgment-guard

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

FAQPage Schema

How do I calibrate AI advice confidence to match evidence strength?▼

Calibrating AI advice confidence involves evaluating evidence strength and recency, surfacing hidden assumptions, and rewriting the recommendation so its expressed certainty matches what the supporting evidence actually justifies on consequential decisions.

What is evidence-to-confidence calibration for high-stakes decision support?▼

Evidence-to-confidence calibration is a decision audit mechanism that grades evidence quality on consequential topics like money, health, and legal issues, then adjusts the recommendation's expressed certainty to prevent overconfidence and surfaces hidden assumptions.

How do I surface hidden assumptions in AI-generated recommendations?▼

You surface hidden assumptions in AI recommendations by mapping fact vs inference, making unspoken premises explicit so users can evaluate the reasoning. This assumption mapping is integrated into a structured decision-audit style output for verification.

Can I use confidence calibration for comparisons and prioritization on consequential topics?▼

Yes, confidence calibration applies to recommendations, comparisons, and prioritization on consequential topics such as money, health, career, legal, safety, and purchases. It applies different rewriting strategies based on whether the evidence scenario is strong, mixed, or weak.

What are the limitations of using evidence grading for risk-aware recommendations?▼

Limitations of evidence grading include its reliance on evaluating epistemic authority and detecting false consensus; the calibrated output remains constrained by available evidence quality and cannot manufacture certainty beyond what the supporting evidence actually justifies.