Anti-Defensive Writing

Anti-Defensive Writing is an open-source academic writing Skill by GitHub user Adkid-Zephyr, in Chinese and English versions. The author calls papers full of "unfortunately" and "limited gains" defensive writing, which hands reviewers the knife in advance. Its principle is to treat a paper as a launch event that makes the case for your single strongest advantage. Twelve rules cover narrative, language, experiments and structure; for example, every experiment must do a job in the argument, and the conclusion only reinforces what readers should remember. In the README's example, "we must acknowledge that we have not yet tested cross-domain generalization" becomes the plainer "cross-domain generalization has not been evaluated." What sets it apart is a complete way of writing that defends less and claims more, with prompts you can copy and use right away.

Who it's for:researchers and academic editors working on papers, abstracts, introductions and rebuttals.

Before you start: the simplest way is to paste the prompt from the repository into a chat. You can also copy the Chinese or English Skill folder into your skills directory, or install the Codex plugin as the README describes. The repository's privacy note says the plugin itself makes no network requests and has no account or telemetry; your paper is processed by the AI you use. Some of its rules are debatable: don't set up a contest on metrics you can't win, don't call a weaker result a loss, and downplay or move experiments that don't support the main story. Used carelessly, these can hide real limitations. It changes the writing only; it doesn't run experiments or check conclusions, and you still need to follow your venue's rules on disclosing limitations and AI use. MIT license.

Based on the project's public documentation. Not tested by AI Inspiration World.