makinuxmakinuxCommunity·1 Agent Skills Included

adversarial-panel

Multi-model adversarial debate for high-stakes decisions

Runs structured multi-round debates where independent models answer a question, attack each other's claims, and defend or concede under fire. Eliminates confidently wrong answers, single-model blind spots, and shallow self-review that plagues important decisions. Delivers one synthesized conclusion with agreements, adjudicated disagreements, calibrated confidence, and falsification conditions.
npx skills add makinux/adversarial-panel --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install adversarial-panel?â–¼

Run `npx skills add makinux/adversarial-panel --all -g -y` in your terminal to install the skill globally.

What is adversarial-panel used for?â–¼

It runs a multi-model adversarial review where independent models answer a question, critique each other across rounds, and a facilitator synthesizes one calibrated conclusion. Use it for consequential, contested, or verifiable questions like architecture decisions and root-cause hypotheses.

Why is multi-model review better than self-critique?â–¼

A single model shares the same blind spots when generating and reviewing its own answer, and tends to rate its own output too highly. Independent models from different families have decorrelated errors, so their critiques catch what one model alone would miss.

Does adversarial-panel work with Claude Code?â–¼

Yes. It is built as a standard SKILL.md file for Claude Code and can recruit panelists from other Claude models or external CLI models like Codex.

When should I trigger an adversarial review?â–¼

Trigger it for high-stakes questions that are contested or verifiable, or whenever you challenge an answer with phrases like 'are you sure?', 'red-team this', or 'get a second opinion'.

Related Repositories in Product & Management

View All in Product & Management→