Orchestrate structured adversarial debates between AI agents with blinded scoring.
npx skills add https://github.com/EZotoff/ez-omo-config --skill debate-ezotoff
Please help me install this Agent Skill. Skill: debate Source: https://github.com/EZotoff/ez-omo-config/tree/main/skills/debate Command: npx skills add https://github.com/EZotoff/ez-omo-config --skill debate-ezotoff
This Skill provides a reproducible, bias-resistant protocol for running structured adversarial debates between AI agents so teams can surface hidden assumptions, stress-test arguments, and reach evidence-based decisions without human bias in scoring.
Start a formal three-round blinded debate with three judges and persist all judge outputs to .sisyphus/debates/{topic}.
High-intent search queries and answers about installing and using this skill.
To run a blinded multi-agent debate, this Skill orchestrates participants under Alpha/Beta labels with deterministic assignment rules, enforcing a segmented S1–S4 structure with per-segment 1–10 scoring to prevent judge bias and reach evidence-based technical decisions.
The best way to automate red team analysis and pre-mortems is using structured adversarial orchestration, which applies sequential round workflows and blinded judging to stress-test arguments and surface hidden assumptions without human scoring bias.
Adversarial analysis handles judge failures through a persist-on-collect reliability mechanism that immediately writes raw judge outputs to disk before parsing, applying malformed-output handling alongside retry and quorum recovery policies to ensure auditable transcripts.
Yes, multi-agent orchestration supports multiple modes including quick 1:1 challenges, allowing you to evaluate technical positions without the overhead of full formal debates, sequential rounds, or a multi-judge quorum.
You should avoid blinded debate orchestration when decisions require subjective human intuition rather than evidence-based technical evaluation, as the Skill enforces mechanical recomputation of scoring totals and rigid segment structures.
Yes, multi-agent debate orchestration automatically persists raw judge background outputs to disk upon collection, storing them in a designated directory path to produce auditable transcripts and final verdicts for architecture reviews.
ajbmachonOrchestrate a three-round multi-agent council debate with visible transcripts.
luicabref97Coordinates scored debates between AI judges to reach a consensus on submissions.
Destynova2Generate Byzantine-fault-tolerant orchestration using the REC-Quorum pattern with adaptive quorum sizing.
pluginagentmarketplaceOrchestrate automated red-team tests across PyRIT, garak, Counterfit, ART, and TextAttack.
