evidence-tournament

$npx mdskill add yogsoth-ai/de-anthropocentric-research-engine/evidence-tournament

Gathers, challenges, and scores evidence for structured debate

  • Solves the problem of verifying claims with reliable and relevant evidence
  • Uses subagents like evidence-scout, debate-critic, and debate-judge
  • Evaluates evidence based on relevance, reliability, recency, and specificity
  • Produces a ranked bracket of evidence quality for decision-making

SKILL.md

.github/skills/evidence-tournamentView on GitHub ↗
---
name: evidence-tournament
description: "Tactic: Evidence gathering, cross-examination, and quality judgment. External evidence is collected, presented, challenged, and scored for relevance and reliability."
type: tactic
used-by: [multiagent-debate]
strategies: [courtroom-structured, critic-defender-judge]
---

# Evidence Tournament Tactic

Structured evidence competition — gather, present, cross-examine, and judge evidence quality.

## Orchestration

1. **evidence-scout** searches for evidence supporting the artifact's claims
2. **evidence-scout** searches for evidence opposing the artifact's claims
3. **debate-critic** presents opposing evidence as structured arguments
4. **debate-defender** presents supporting evidence as counter-arguments
5. **cross-examination** probes both evidence sets:
   - Source reliability assessment
   - Relevance to specific claims
   - Recency and applicability
   - Potential confounds or alternative interpretations
6. **debate-judge** scores each piece of evidence and produces tournament bracket result

## Evidence Quality Criteria

- **Relevance**: Direct bearing on the claim (0.0–1.0)
- **Reliability**: Source credibility and methodology (0.0–1.0)
- **Recency**: Temporal applicability (0.0–1.0)
- **Specificity**: How precisely it addresses the claim (0.0–1.0)

## Subagents Dispatched

- evidence-scout × 2 (pro and con evidence gathering)
- debate-critic (opposing evidence presentation)
- debate-defender (supporting evidence presentation)
- cross-examination (evidence probing)
- debate-judge (evidence scoring and verdict)

## Termination Conditions

- All claims have been evidenced and cross-examined
- Evidence search budget exhausted (2/5/10 searches)
- No further relevant evidence discoverable
- Judge has scored all evidence pairs

More from yogsoth-ai/de-anthropocentric-research-engine