Evidence Evaluator Skill

Purpose

Assess source credibility and evidence quality using CRAAP test and evidence hierarchies.

Trigger Conditions

Keywords: "evaluate source", "is this credible", "check evidence", "study quality", "reliable source"

Procedure

Phase 1: CRAAP Test

C - Currency: How recent? R - Relevance: Fits the need? A - Authority: Qualified author? Reputable publisher? A - Accuracy: Verifiable? Sources cited? Peer-reviewed? P - Purpose: Why created? Bias present?

Score each: 0-5 (Weight: R=1.5x, A=2x, A=2x, others=1x) Total score / 37.5 = Credibility %

Phase 2: Evidence Hierarchy

Rank by epistemic strength:

1. Systematic reviews & meta-analyses (Highest)
2. Randomized controlled trials
3. Cohort studies
4. Case-control studies
5. Cross-sectional studies
6. Case reports
7. Expert opinion
8. Anecdote (Lowest)

Phase 3: Study Quality (if applicable)

• Design: Appropriate for question?
• Sample size: Adequate power?
• Bias: Selection, measurement, reporting?
• Confounds: Controlled?
• Statistics: P-values, effect sizes, CIs?
• Reproducibility: Replicable?

Phase 4: Verdict

• Credibility: 90-100% Highly credible / 70-89% Credible / 50-69% Questionable / <50% Not credible
• Use recommendation: ✅ Use / ✓ Use with caveats / ⚠️ Verify first / ❌ Don't use

Tool Access

Allowed: Read

Integration

Agent: evidence-evaluator Protocols: THINK-EVIDENCE-001 to 005

Example

Input: Blog post claiming "Coffee causes cancer"

Output:

• CRAAP: C=3, R=3, A=1 (no qualifications), A=1 (no sources), P=2 (clickbait) = 25% credible
• Hierarchy: Anecdote (lowest level)
• Actual evidence: WHO meta-analyses show coffee protective against several cancers
• Verdict: ❌ Don't use blog post, use systematic reviews instead