# Calibrate Confidence

> Match certainty to evidence strength

- Skill: `synapseradio/calibrate-confidence` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add synapseradio/calibrate-confidence`
- Raw SKILL.md: https://api.skillmd.com/api/skills/synapseradio/calibrate-confidence/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: synapseradio (https://skillmd.com/u/synapseradio)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/synapseradio/calibrate-confidence

---


Follow these steps:

### 1. Identify the claim

Target: `$ARGUMENTS`, else recent assertions in the current context.

### 2. Name confidence explicitly

State the certainty level in concrete terms, distinguishing components if it varies:

- "Very confident (90%+), because…"
- "Moderately confident (60–80%), because…"
- "Uncertain (40–60%), due to…"
- "Low confidence (below 40%), because…"

### 3. Inventory the evidence

What does the confidence rest on?

- Direct experience or observation
- Documentation or authoritative sources
- Logical reasoning from established facts
- Inference or analogy
- Assumption without evidence

Assess how much is evidence-backed versus assumption.

### 4. Check for overconfidence

Definitive statements on limited data; uncertainty suppressed to sound authoritative; "obviously" / "clearly" without demonstration.

### 5. Check for underconfidence

Good evidence discounted; hedging from habit rather than need; excessive qualification of well-supported claims.

### 6. Test calibration

What odds would you accept betting on this? What evidence would change the level? Does past track record suggest over- or underconfidence?

### 7. Communicate uncertainty

Express confidence in proportion to evidence. Distinguish personal ignorance from collective uncertainty. Show where certainty ends and speculation begins.