What it does

Immortal-skill is a prompt-and-protocol framework for “distilling” people from chat logs, documents, and public content into structured AI personas. It supports 12+ platforms (WeChat, Feishu, iMessage, Telegram), offers seven role templates, and claims alignment with something called the OpenClaw Soul Spec. The core engine lives in SKILL.md; surrounding repos handle public-figure extraction (steamer-skill), anti-distillation defenses (distill-shield-skill), and usage rights (distill-protocol-skill).

The interesting bit

The project treats persona extraction as a four-dimensional distillation—not just what someone said, but how they think, decide, and speak. The “steamer” sub-project is the grabby part: it argues that since you paid for Elon Musk’s books, Sam Altman’s blog posts, and Warren Buffett’s shareholder letters, you’re entitled to repurpose their validated cognitive frameworks as personal AI advisors. There’s even a prefab “persona plaza” with distilled Musks, Einsteins, and Cleopatras.

Key highlights

• 12+ platform ingestion including WeChat, Feishu, iMessage, Telegram
• Seven role templates for different relationship types (self, colleague, mentor, family, etc.)
• Four-dimensional distillation targeting cognition, speech patterns, decision style, not just content
• Ecosystem approach: separate repos for extraction, legal-ish rights separation, and anti-distillation “poisoning” defenses
• Public figure prefabs available via agenworld.com/market

Caveats

• The “OpenClaw Soul Spec” and “AgentSkills Standard” appear to be project-defined or niche standards; no independent verification visible
• Actual implementation quality is unclear—much of the repo appears to be markdown specifications and prompt templates rather than executable code
• The “cow protection law” (牛马保护法) is explicitly noted as a joke-with-serious-underpinnings, not actual legal framework

Verdict

Worth a look if you’re building persona-extraction pipelines or exploring the ethics of digital identity. Skip if you need battle-tested, production-ready tooling—this is more conceptual framework than shipped software.