---
name: yogsoth-ai/evidence-weighing
source: https://app.decimal.ai/s/yogsoth-ai-evidence-weighing@1/SKILL.md
source_sha256: d09ebd4de90e
---

# Evidence Weighing

Assess the strength of evidence for causal claims. Not all evidence is equal — RCTs > observational studies > expert opinion > anecdote.

## Available SOPs

- evidence-linking — connect evidence to claims
- confidence-scoring — assign calibrated confidence
- contradiction-flagging — identify conflicting evidence

## Guiding Principles

- **Hierarchy of evidence.** Interventional > observational > theoretical > anecdotal.
- **Effect size matters.** A statistically significant but tiny effect is weak evidence for practical causation.
- **Replication is king.** Single studies are suggestive; replicated findings are convincing.
- **Confounders weaken.** Evidence from studies with uncontrolled confounders gets lower weight.
- **Mechanism strengthens.** Evidence backed by a plausible mechanism gets higher weight.

## Minimum Yield

<HARD-GATE>
≥2 evidence assessments with explicit strength ratings per invocation.
</HARD-GATE>

<!-- BEGIN available-tables (generated) -->

## Available SOPs

Optional, no fixed order; the final leaf is always a sop.

| SOP | When to use |
| --- | --- |
| confidence-scoring | SOP for assigning calibrated confidence scores to causal claims based on evidence quality and quantity. |
| contradiction-flagging | SOP for flagging contradictions in the causal model — identify conflicting evidence or mechanism claims. |
| evidence-linking | SOP for linking evidence pages to causal claims — creates supported_by or contradicts edges. |

<!-- END available-tables (generated) -->