kill-argument - Adversarial attack-defense review for theory papers
Runs a fresh two-thread review that writes a strongest rejection memo, defends it point by point, and identifies unresolved critical issues.
Tags
Updated: 2026-10-19Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Discover paper source files
- Generate rejection memo
- Decompose rejection points
- Defend rejection points
- Classify unresolved issues
- Write review reports
- Render reports to HTML
Inputs
- Paper directory
- LaTeX source files
- Bibliography files
- Figure files
- Compiled PDF
- HTML rendering setting
- Reviewer model selection
Outputs
- KILL_ARGUMENT.md
- KILL_ARGUMENT.json
- Rendered HTML report
- Adversarial accept/reject verdict
Requirements
- Bash and file access
- mcp__codex__codex access
- Supported Codex reviewer model
- Read-only reviewer sandbox