verdict-synthesis
$
npx mdskill add yogsoth-ai/de-anthropocentric-research-engine/verdict-synthesisSynthesizes campaign findings into typed verdict reports.
SKILL.md
.github/skills/verdict-synthesisView on GitHub ↗
--- name: verdict-synthesis description: Synthesizes findings from a completed campaign into typed verdict reports. Produces DebateVerdict, RedTeamReport, FailureAnticipationReport, CounterfactualMap, or AdversarialStressReport depending on campaign. Also supports cross-campaign StressTestSummary. execution: subagent prompt: ./prompt.md input: campaign_name (string), strategy_outputs (string), weakness_classifications (string) used-by: multiagent-debate, red-teaming, failure-anticipation, counterfactual-probing, adversarial-stress-testing --- # Verdict Synthesis Synthesizes campaign findings into typed verdict reports. ## Execution Subagent — spawned via subagent-spawning/spawn-agent. ## Why Subagent Synthesis requires processing all strategy/tactic outputs from a campaign in dedicated context to produce coherent final report. ## Input - **campaign_name**: Which campaign produced these findings - **strategy_outputs**: All strategy and tactic outputs from the campaign - **weakness_classifications**: Classified weaknesses (from weakness-classification) ## Output Typed report matching campaign: - `DebateVerdict` (multiagent-debate) - `RedTeamReport` (red-teaming) - `FailureAnticipationReport` (failure-anticipation) - `CounterfactualMap` (counterfactual-probing) - `AdversarialStressReport` (adversarial-stress-testing) - `StressTestSummary` (cross-campaign aggregation) ## Budget One unit = one synthesis pass producing a complete typed report.