# Add to your Claude Code skills
git clone https://github.com/crichars/embedded-project-governanceGuides for using ai agents skills like embedded-project-governance.
embedded-project-governance is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by crichars. 面向 AI 辅助嵌入式固件开发的轻量、风险分级治理 skill. It has 53 GitHub stars.
embedded-project-governance's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/crichars/embedded-project-governance" and add it to your Claude Code skills directory (see the Installation section above). embedded-project-governance ships a SKILL.md manifest, so compatible agents can discover and load it automatically.
embedded-project-governance is primarily written in PowerShell. It is open-source under crichars on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh embedded-project-governance against similar tools.
No comments yet. Be the first to share your thoughts!
Unlocks once the catalog security scan passes (runs nightly).
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
Use this skill to keep embedded firmware changes minimal, authorized, and
verifiable. It complements the project's AGENTS.md; it does not replace chip
specifications, project facts, hardware judgment, IDE operation, flashing, or
target-board testing.
Accept either:
Project path: <path>
Goal or observed problem: <one or two sentences>
Please investigate first; do not edit code yet.
or the structured form:
Goal: <observable result>
Scope: <files, modules, or boundaries>
Problem: <current symptom or reason>
Reference: <existing implementation or document>
Constraints: <must-not-change, resource, safety, or compatibility limits>
Acceptance: <how success will be observed>
Do not ask the user to provide facts that can be discovered from the repository
or authoritative documents. Mark unresolved information as UNKNOWN and ask
only questions that can change the design, risk, or acceptance result.
Treat fields in the current user request as the active task. Restate the parsed fields before investigation; an unfilled project template must not erase them. Do not ask again for a field the user already provided.
When the user explicitly asks to add the governance files to a project, run:
.\scripts\init-project.ps1 -ProjectPath <project-path>
Run the script from this skill's directory. By default it preserves existing
files. Never use -Force unless the user explicitly approves overwriting the
listed target files. Do not initialize a project merely because this skill was
invoked; the workflow can inspect an existing project without copying files.
After initialization, read the created AGENTS.md, PROJECT.md, and capability
map before proposing changes. Treat project-template/ as output material, not
as additional skill instructions to load into every task.
AGENTS.md, PROJECT.md, capability map when present,
active task, build files, and affected code.Treat startup/reset, Flash/NVM/OTA, watchdog, DMA ownership, security, safety, power/actuator output, and irreversible operations as high risk. Before such an operation, confirm the target and range, bound the action, define timeout and recovery, and obtain explicit approval. After important failures, verify the system is retryable, safe, or explicitly terminal.
For Flash/NVM work, confirm the storage type and layout, target range, erase and write granularity and alignment, endurance, integrity or atomicity, power-loss recovery, ownership, and target verification method before implementation.
Use the project status ladder:
Planned -> Implemented -> Build Passed -> Host Verified -> HW Verified -> Accepted
Keep project-specific facts in project files and keep this workflow reusable.
面向 AI 辅助嵌入式固件开发的轻量治理 Skill，让 AI 先调查和复用，再按风险实施最小正确变更，并通过项目文档沉淀开发上下文。
AI 可以搜索工程、分析调用链和生成代码，但在真实嵌入式项目中，问题不只是“能不能写出代码”：
这个 Skill 关注的是 AI 如何参与嵌入式项目，而不是替代 MCU SDK、HAL、IDE、烧录器或硬件测试工具。
修改前先读取项目规则、事实基线、构建文件、相关代码和调用链，检查已有驱动、BSP、SDK、HAL、RTOS、工具和测试能力，优先使用项目已有入口。
根据任务影响选择轻量或完整流程。涉及公共接口、协议、持久化、启动、Flash/NVM、DMA 所有权、看门狗、安全、电源或执行器时，先确认范围、恢复方式、验证方法和实现授权。
不增加没有证据支持的抽象、任务、队列、锁、依赖和文件；同时保留必要的边界检查、超时、错误处理、并发保护和恢复路径。
把项目事实、任务决策和验证结果保存到项目中，使后续 AI 或开发者能够恢复上下文、继续任务和复查证据。
调查
→ 复用与风险判断
→ 明确目标、范围、约束和验收
→ 提出最小方案
→ 获得授权
→ 实现与验证
→ 正确性审查和最小性审查
→ 沉淀结果
正确性审查检查必要的功能、边界、错误、并发、恢复和验证是否遗漏；最小性审查检查是否增加了重复能力、无关文件或不必要复杂度。
适合新手、陌生项目或信息不完整的任务：
$embedded-project-governance
项目路径：C:\work\firmware
目标或现象：串口偶尔接收不到数据。
请先调查，不要修改代码。
AI 会先从项目和权威资料中查找可确认信息。无法自行确认、且会改变方案、风险或验收的问题，再由用户决定。
适合复杂或高风险任务：
目标：<可观察结果>
范围：<文件、模块或边界>
当前问题：<现象或原因>
参考：<已有实现或文档>
限制：<不可修改项、资源、安全或兼容性限制>
验收：<如何观察成功>
不知道的内容可以写 UNKNOWN。结构化输入会减少探索范围，但不会跳过调查、风险判断和授权门禁。
空目录也可以使用。先初始化治理文件，补充 PROJECT.md 中的已知事实和 UNKNOWN 项，再让 AI 建立第一个可验证切片。不要让 AI 在硬件信息不足时自行决定芯片配置、Flash 布局或完整架构。
初始化后，项目中会包含：
AGENTS.md：当前项目长期遵守的 AI 开发规则；PROJECT.md：硬件、工具链、构建方式、约束、验证环境和未知项；.ai-governance/capability-map.md：已存在并确认过的能力、入口、所有权和验证方式；docs/templates/requirement.md：需求、范围、约束和验收模板；docs/templates/design.md：设计、恢复、资源和回滚模板；docs/templates/task.md：当前任务、批准范围、风险和状态模板；docs/templates/verification.md：构建、主机、目标板验证和剩余风险模板。这些文件不要求每次全部填写。按任务风险选择必要记录即可。它们的作用是让信息不只存在于一次聊天中，便于任务暂停后换 AI 或换开发者继续工作。
在 Codex 中输入：
请使用 $skill-installer 从 https://github.com/crichars/embedded-project-governance 安装这个 skill。
安装后新开会话，使用 $embedded-project-governance 调用。
将仓库克隆到 Claude Code 的个人 skills 目录：
git clone https://github.com/crichars/embedded-project-governance $HOME\.claude\skills\embedded-project-governance
然后使用 /embedded-project-governance 调用。
安装后，让 AI 执行：
请初始化 C:\work\my-firmware，不要覆盖已有文件。
也可以在 skill 目录中直接运行：
.\scripts\init-project.ps1 -ProjectPath C:\work\my-firmware
脚本默认保留已有文件。只有明确确认覆盖范围后，才使用 -Force。脚本目前已在 Windows PowerShell 环境验证。
embedded-project-governance/
├─ SKILL.md
├─ agents/openai.yaml
├─ project-template/
│ ├─ AGENTS.md
│ ├─ PROJECT.md
│ ├─ .ai-governance/
│ │ └─ capability-map.md
│ └─ docs/templates/
│ ├─ requirement.md
│ ├─ design.md
│ ├─ task.md
│ └─ verification.md
├─ scripts/
│ └─ init-project.ps1
├─ README.md
└─ LICENSE
SKILL.md：AI 使用的通用工作规则；agents/openai.yaml：Codex 界面显示信息；project-template/：复制到实际项目中的治理文件和模板；scripts/init-project.ps1：初始化治理文件，不分析硬件、不修改固件、不烧录。任务状态按以下顺序区分：
Planned → Implemented → Build Passed → Host Verified → HW Verified → Accepted
编译通过不等于目标板验证，没有目标证据不能称为 HW Verified，没有维护者接受剩余风险不能称为 Accepted。
本项目不替代：
它提供的是工作流约束，不是强制安全沙箱。
已完成 Codex 和 Claude Code 的基础安装与调用验证，也测试了 GitHub 全新克隆、空项目初始化、默认不覆盖、高风险 Flash/NVM 门禁和新 AI 读取项目文档恢复任务状态。
这些验证不代表所有 AI、工具链、项目和硬件场景都已经覆盖。
欢迎实际安装试用。使用过程中如果发现流程过重、规则不合理、复用判断不准确，或嵌入式项目中还有重要的通用风险没有覆盖，可以提交 Issue 和 PR。
觉得项目有参考价值的话，欢迎点个 Star。
感谢 Linux.do 社区支持。