argue

AI debate resolves complex topics by pitting two models against each other to argue opposing positions. This Skill runs an adversarial Claude-vs-Gemini debate with convergence checks, generating a final synthesis saved to ~/.claude/arguments/finals/<slug>.md.

Yes, adversarial AI debates support optional context files and seeds. The workflow parses your topic and optional flags, loads the provided context, and generates independent opening positions before starting the back-and-forth argumentation.

Adversarial debate synthesis works with plan, bug, tech, or general topics. The workflow applies multi-step argument parsing, opening positions, back-and-forth debate, and convergence checks to produce a structured final document.

Debate convergence works by conducting a structured back-and-forth exchange between models with periodic convergence checks. Once the arguments align, the workflow generates a final synthesis and writes it to the finals directory with optional history transcripts for traceability.

No external dependencies are required to run the debate workflow. The Skill operates independently using the existing Claude environment and its built-in model integration to execute the adversarial argument parsing and synthesis generation.

Yes, debate history transcripts are optionally saved alongside the final synthesis. The workflow outputs the decisive conclusion to the finals directory and can store the full argumentation history, ensuring complete traceability of the back-and-forth debate process.