可重放、可评测、可分叉的多 Agent 对抗社会 Harness;狼人杀是首个领域适配器,React Cockpit 提供运行与证据分析。
-
Updated
Jul 24, 2026 - TypeScript
可重放、可评测、可分叉的多 Agent 对抗社会 Harness;狼人杀是首个领域适配器,React Cockpit 提供运行与证据分析。
GAN-inspired adversarial development loop — Planner directs Generator + Evaluator in fresh cmux panes, gated by hard pass/fail per deliverable. Adapted from Anthropic's harness-design guidance for long-running agentic apps.
Stage adversarial Claude debates: two counsel agents argue, a judge rules
Adversarial multi-agent NDA negotiator. Buyer vs Seller AI agents with private mandates, semantic convergence detection, and redlined DOCX output. AutoGen · Claude API.
Adversarial multi-agent investment due-diligence. Milan AI Week / lablab.ai hackathon. SEC EDGAR + earnings call audio + Bull/Bear/Reconciler agents on Gemini 2.5 Pro + Featherless Qwen3 + Speechmatics, hosted on Vultr.
Add a description, image, and links to the adversarial-agents topic page so that developers can more easily learn about it.
To associate your repository with the adversarial-agents topic, visit your repo's landing page and select "manage topics."