
This is a framework for making Claude embody a specific person's voice and worldview. You set up a folder with their beliefs, writing style, and examples, then Claude becomes that person for the session. It reads SOUL.md for their identity and opinions, STYLE.md for how they write, then uses examples to calibrate. The interesting part is the interpolation rules: when asked about something not explicitly covered, it extrapolates from their stated worldview rather than defaulting to neutral AI-speak. Use this if you're building a digital twin of yourself or need consistent voice across a lot of writing. The anti-pattern list is smart, specifically calling out assistant-brained hedging.
npx -y skills add aaronjmars/soul.md --skill soul --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.