Nuwa-skill is an Agent Skills protocol implementation that distills the thinking patterns of notable figures to augment AI agent reasoning.
The tool addresses the problem of capturing and reusing how specific people think—their mental models, decision heuristics, and communication patterns—rather than just their surface-level habits or documented processes. It works by extracting five layers of cognition: how someone speaks (expression DNA), how they think (mental models and frameworks), how they judge (decision heuristics), what they avoid (anti-patterns and values), and where they acknowledge limitations. Once distilled, these patterns can be invoked within an AI agent to analyze problems through that person's cognitive framework rather than simply mimicking their words.
The tool suits developers working within Agent Skills-compatible runtimes who want to leverage specific thinking styles for problem-solving. It works across more than fifty compatible environments including Claude Code, Cursor, Codex, OpenClaw, and others. The README positions it as an alternative to distilling colleagues, arguing instead for distilling figures like Jobs, Musk, Munger, or Feynman whose cognitive frameworks offer broader applicability. Installation is straightforward through a single command, manual directory placement, or by pasting the skill's markdown content directly into agent conversations.
The project shows active development with multiple language translations of its documentation, indicating sustained effort to reach different developer communities. The codebase includes tooling for specific runtime integrations, such as dedicated installation scripts for particular platforms. The README demonstrates concrete usage examples showing how distilled thinking patterns produce analysis grounded in specific mental models rather than generic responses.
