szl-cookbook-source
收藏资源简介：
SZL Cookbook — Source Mirror 是 GitHub 仓库 `szl-holdings/szl-cookbook` 在 Hugging Face 平台上的源代码镜像数据集，旨在提升该资源在开源社区的可发现性和可访问性。该数据集是 SZL Holdings 研究栈的组成部分，与其正式验证论文、AI对齐模型等构件紧密关联。核心内容是一套工程化食谱，提供了在 SZL 基底上构建受治理人工智能系统的 9 个可组合技能指南，涵盖预运行推理、代码重构、审查、调试、依赖管理、死代码处理、文档编写和提交规范等领域。数据集包含源代码文件，并确保每个声明均可追溯至 Zenodo DOI、GitHub 提交哈希，并在适用时关联基于 Mathlib v4.13.0 的 Lean 4 形式化证明。这是一个在特定提交时间点（2026-05-29）的快照，并非实时更新源，且为精简体积，排除了 `.git/` 目录、`node_modules/` 及大于 50MB 的二进制文件。根据元数据，其规模小于 1000 个样本，语言为英语。该数据集主要服务于对 AI 治理、形式化方法、软件供应链安全（如 SLSA、DSSE）感兴趣的研究人员和开发者，作为参考、验证和复现的源代码库，而非用于机器学习模型训练。数据集采用 Apache-2.0 许可证发布，作者为 Stephen P. Lutar。
SZL Cookbook — Source Mirror is a source code mirror dataset of the GitHub repository `szl-holdings/szl-cookbook` on the Hugging Face platform, aimed at enhancing the discoverability and accessibility of this resource in the open-source community. This dataset is part of the SZL Holdings research stack and is closely linked to its formal verification papers, AI alignment models, and other components. The core content is a set of engineering recipes, providing 9 composable skill guides for building governed AI systems on the SZL foundation, covering areas such as pre-run reasoning, code refactoring, review, debugging, dependency management, dead code handling, documentation writing, and commit specifications. The dataset includes source code files, ensuring that each claim can be traced back to Zenodo DOI, GitHub commit hashes, and, where applicable, linked to Lean 4 formal proofs based on Mathlib v4.13.0. This is a snapshot at a specific commit time (2026-05-29), not a live updated source, and to reduce size, it excludes the `.git/` directory, `node_modules/`, and binary files larger than 50MB. According to metadata, it has a scale of less than 1000 samples and is in English. The dataset primarily serves researchers and developers interested in AI governance, formal methods, and software supply chain security (e.g., SLSA, DSSE) as a source code repository for reference, verification, and reproduction, rather than for machine learning model training. It is released under the Apache-2.0 license, authored by Stephen P. Lutar.
数据集概述：SZL Cookbook — Governed AI Engineering Recipes
- 数据集名称: SZL Cookbook — Governed AI Engineering Recipes
- 许可证: Apache-2.0
- 数据集规模: n < 1K
- 任务类别: 其他（other）
- 语言: 英语（en）
- 标签: cookbook, agentic-ai, mcp, governance, dsse, slsa, anthropic, patterns
内容简介
该数据集是 github.com/szl-holdings/szl-cookbook 的镜像仓库，包含 9 个 SKILL.md 文件，用于构建基于 SZL 基座的受控 AI 系统。模式覆盖了 Anthropic MCP 模式、DSSE 收据集成以及 OTel 跨度发射。
关键信号
- 配方模式数量: 9 个 SKILL.md 文件
- MCP 模式: 兼容 Anthropic
- 许可证: Apache-2.0
内容列表
| 配方名称 | 描述 |
|---|---|
| SKILL_01 | MCP 收据发射模式 |
| SKILL_02 | DSSE 信封构建 |
| SKILL_03 | OTel 跨度附加 |
| SKILL_04 | Λ 轴策略门 |
| SKILL_05 | SLSA 证明 |
| SKILL_06 | Bekenstein 上下文预算 |
| SKILL_07 | PAC-Bayes 稳定性检查 |
| SKILL_08 | Reidemeister 审计收尾 |
| SKILL_09 | 跨组件收据链 |
相关链接
- 运行器: SZLHOLDINGS/szl-cookbook-runner
- 平台: SZLHOLDINGS/szl-cookbook-platform
- 源代码: github.com/szl-holdings/szl-cookbook
引用格式（BibTeX）
bibtex @misc{lutar2026ouroboros, title = {Ouroboros: Formal Verification of Agentic AI Governance — v18.0}, author = {Lutar, Stephen P.}, year = {2026}, doi = {10.5281/zenodo.20434276}, url = {https://doi.org/10.5281/zenodo.20434276} }
联系方式
- 作者: Stephen P. Lutar
- 邮箱: stephen@szlholdings.com
- ORCID: 0009-0001-0110-4173
- 相关主页: github.com/szl-holdings · huggingface.co/SZLHOLDINGS
