An open-source LLM skill for detecting and rewriting defensive academic writing while preserving necessary scientific caveats and evidence boundaries.
