Agent skill

confidence-scoring

Assess research quality with systematic confidence scoring, source validation, and uncertainty quantification.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/other/confidence-scoring

SKILL.md

Research Confidence Scoring

Overview

Apply systematic confidence scoring to research findings. Quantify uncertainty, validate sources, detect contradictions, and provide calibrated assessments of information reliability.

The Process

1. Source Assessment

Rate each source on a 0.0-1.0 scale:

Score Source Quality
0.9-1.0 Primary source, official docs, peer-reviewed
0.7-0.8 Reputable secondary source, established experts
0.5-0.6 Community content, blog posts, forums
0.3-0.4 Unverified claims, single source only
0.0-0.2 Contradicted by better sources, outdated

Source Quality Factors:

  • Authority: Is the source authoritative on this topic?
  • Recency: How current is the information?
  • Corroboration: Do multiple sources agree?
  • Methodology: Is reasoning/evidence provided?
  • Bias: Does the source have obvious bias?

2. Claim Validation

For each research finding:

Claim: [Statement being assessed]
Sources: [List of sources]
Source Confidence: [0.0-1.0 per source]
Corroboration: [Number of independent sources agreeing]
Contradictions: [Sources that disagree]
Final Confidence: [Weighted score]

Confidence Calculation:

  • Single source: Cap at 0.6 regardless of source quality
  • Two agreeing sources: Up to 0.8
  • Three+ agreeing sources: Up to 0.95
  • Any contradiction: Reduce by 0.2, flag for review

3. Uncertainty Quantification

Express findings with explicit uncertainty:

High Confidence (0.8-1.0):

"X is documented behavior" / "X is confirmed by multiple sources"

Medium Confidence (0.5-0.7):

"X appears to be the case based on..." / "Evidence suggests X"

Low Confidence (0.2-0.4):

"X may be true, but verification needed" / "Limited evidence for X"

Uncertain (0.0-0.2):

"Cannot confirm X" / "Conflicting information about X"

4. Contradiction Detection

When sources disagree:

  1. Identify the contradiction: State both positions clearly
  2. Assess source quality: Which sources are more authoritative?
  3. Check recency: Is one source outdated?
  4. Look for context: Could both be true in different contexts?
  5. Flag unresolved: If contradiction persists, report both positions

5. Confidence Reporting

Structure research output with confidence metadata:

markdown
## Finding: [Topic]

**Confidence: 0.75** (Medium-High)

**Summary:** [Key finding in 1-2 sentences]

**Evidence:**
- Source A (0.9): [What it says]
- Source B (0.7): [What it says]
- Source C (0.5): [What it says]

**Limitations:**
- [What we couldn't verify]
- [Areas of uncertainty]

**Contradictions:**
- [If any sources disagreed]

Confidence Calibration

Red Flags (Lower confidence)

  • Only one source found
  • Source is promotional/marketing content
  • Information is > 2 years old for fast-moving topics
  • Source has obvious bias or conflict of interest
  • Claim seems extraordinary without extraordinary evidence

Green Flags (Higher confidence)

  • Multiple independent sources agree
  • Official documentation confirms
  • Peer-reviewed or expert-validated
  • Includes methodology or reasoning
  • Matches observed behavior/testing

Key Principles

  • Explicit uncertainty: Always state confidence level
  • Source transparency: Cite sources with quality ratings
  • Contradiction awareness: Flag disagreements, don't hide them
  • Calibration: Track accuracy to improve future scoring
  • Epistemic humility: "I don't know" is a valid answer
  • Recency weighting: Recent sources > older for evolving topics

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results