claim-vs-evidence-checker

Files (2)

• SKILL.md
• openai.yaml · agents/

SKILL.md (first lines)

# Claim vs Evidence Checker

## Goal

Judge how well a specific claim matches the available evidence.

This skill does not only search for papers. It compares the strength of the wording in the claim against the strength, directness, and consistency of the evidence.

## When To Use

- the user asks "is this really supported?"
- a bold statement needs an evidence match check
- scientific-family work needs a final claim-strength pass

## Scope Boundaries

In scope:

- restate the target claim
- compare claim strength to evidence strength
- label the claim as supported, tentative, overstated, or unsupported

Out of scope by default:

- full literature discovery from zero without narrowing
- policy or medical advice beyond the evidence comparison

## Workflow

1. State the exact claim.
2. Identify the relevant evidence set.
3. Compare:
   - claim breadth
   - evidence directness
   - evidence quality
   - consistency
4. Judge the fit between wording and support.
5. Rewrite the claim conservatively if needed.

## Output Contract

1. `target_claim`
2. `evidence_base`
3. `claim_strength`
4. `evidence_strength`
5. `fit_assessment`
6. `conservative_rewrite`
7. `recommended_followup`
8. `next_step`

## Guardrails

- Do not let exciting findings justify stronger wording than the evidence supports.
- Do not treat one supportive study as settled support.
- Prefer conservative restatement over rhetorical overreach.

From Alubiama/awesome-cognitive-skills, MIT.

More from Alubiama/awesome-cognitive-skills