embed-ai-tool skill
24 skills for microcontroller firmware: build, flash, GDB debugging, UART, Modbus, CAN, logic analyzers and static analysis
- 1,068 ★ on the repository
- Last commit
Broad access
Why this level
- The skills flash a board and can erase working firmware on the connected device
- modbus-debug and can-debug write to registers and buses of real equipment, where a command can start a drive or change settings
- The agent runs third-party tools and scripts on your computer
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/embed-ai-toolDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The skills flash a board and can erase working firmware on the connected device
- modbus-debug and can-debug write to registers and buses of real equipment, where a command can start a drive or change settings
- The agent runs third-party tools and scripts on your computer
This is third-party code. Review the repository files before installing.
What it does
embed-ai-tool is a set of twenty-four skills that guides an agent through the whole firmware development cycle for microcontrollers. Builds are covered by skills for CMake, Keil MDK, IAR EWARM, PlatformIO, ESP-IDF and plain Makefiles. Flashing goes through Keil, OpenOCD, PlatformIO, ESP-IDF and SEGGER J-Link, and J-Link can capture an RTT log. Debugging uses skills with GDB over OpenOCD or J-Link, while rtos-debug understands FreeRTOS, RT-Thread and Zephyr threads, shows stack headroom and finds deadlocks. A separate group works with ports and instruments. serial-monitor captures a port log, serial-shell opens an interactive session, modbus-debug reads and writes Modbus RTU and TCP registers, can-debug listens to the bus and sends frames, visa-debug talks to instruments over SCPI and logic-analyzer collects Saleae signals and decodes I2C, SPI, UART and CAN. Analysis comes as two skills: a memory map from .map or ELF files and static analysis through cppcheck, clang-tidy and the GCC analyzer with MISRA-C checks. The workflow skill chains steps into a pipeline, for example build, flash, watch the port. The scripts remember the last successful profile in a .em_skill.json file in the project root and reuse it on the next run.
Who it is for. Embedded developers on STM32, ESP32 and other microcontrollers whose firmware builds in Keil, IAR, CMake or PlatformIO.
Good fit when
- A project builds in Keil or IAR and the agent should run build and flash on its own and read the port log
- Firmware crashes and you need a GDB session over OpenOCD or J-Link with a breakpoint and state inspection
- A FreeRTOS, RT-Thread or Zephyr system is suspected of a deadlock or a stack shortage
- A device needs checking over Modbus or CAN: read registers, scan slaves, listen to the bus
- Before release you need a memory usage report and static analysis with MISRA-C checks
Not a fit when
- The project has nothing to do with microcontrollers: the skills target firmware build and debug
- The attached hardware must not be touched without a person: the skills flash the board and write registers
- You need a single step such as a CMake build: install individual skills rather than the whole set
- Keil, OpenOCD or J-Link are not set up: the skills call installed tools and do not install them
Example request
Build the project in Keil, flash the board and show the first 20 lines of the UART log; if the build fails, stop and explain the errorLimitations
The project description is written in Chinese by default, with an English version in README_EN.md. The skills call external tools (Keil, IAR, OpenOCD, J-Link, PlatformIO, ESP-IDF, GDB) and work only if those are installed and found at the paths from the config. Linux, macOS and Windows are declared. GitHub detected no license for the repository. The embed-ai-tool skill itself is marked internal and exists to pick skills, not to work with a board.
How to disable. Run npx skills remove -g and pick the skills. A script-based install is removed through the install.py options. Delete the .em_skill.json file in the project root and any saved tool paths by hand.
FAQ
How do I install embed-ai-tool?
npx skills add LeoKemp223/embed-ai-tool -g -y installs all skills, and the same command with --skill flags installs a selection, for example --skill build-cmake --skill flash-openocd. There is also a scripted install through scripts/install.py that analyzes the project.
Why does the agent ask questions before installing?
INSTALL.md requires the agent to study the project first, show the build system, debugger and protocols it found, and wait for your choice. Copying the whole skills folder into the project is forbidden so as not to clutter it.
Which build systems are supported?
CMake, Keil MDK, IAR EWARM, PlatformIO, ESP-IDF and plain Makefiles. For flashing there are Keil, OpenOCD, PlatformIO, ESP-IDF and J-Link.
How does a skill learn project parameters on a repeat run?
The build and flash scripts take the last successful profile from the .em_skill.json file in the project root. Explicit parameters win, and the --resume flag only demands that the cache exist.
What if it is unclear which skill is needed?
Say it briefly, for example flash or build. The internal embed-ai-tool skill lists the matching options in that category and you pick one.
Discussion
No questions or reviews yet.
Related
SkillsEmbedded Skills
embeddedskills
Skills for embedded development: AI drives compilers, debuggers (J-Link and others) and buses (CAN) from code to hardware checks. README in Chinese and English
Claude Code skills and a Raspberry Pi testbench: the agent writes ESP32 firmware, flashes the board and tests it on hardware
Dimillian Skills
Skills
A set of 16 engineering skills for agents: Apple platform work, macOS packaging, React performance, code review and refactoring
Nine JS Mastery skills take a change through phases: from idea and spec to verification, tests, release texts and a docs sync
A Verilog-2001 design skill that writes RTL from a brief and explains and repairs existing modules with a check report
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
README in short
The README promises automation of the development, debugging and testing cycle for hardware products and right away tells the agent to read INSTALL.md before installing. Then come a one-phrase quick start, a table of twenty-four skills in six categories, dialogue and command examples, the repository layout and shared contracts: project profile, context handoff between skills, result schema and failure taxonomy.
README badge
Original SKILL.md
--- name: embed-ai-tool description: 嵌入式开发技能集的总控入口。负责两类任务：(1) 用户请求安装本仓库 skill 时，引导选择全部或按需安装；(2) 用户发出模糊指令（如"烧录"、"编译"、"调试"）且无法自动判断应使用哪个具体 skill 时，列出同分类下的候选 skill 供用户选择。 metadata: internal: true --- # embed-ai-tool 总控 本技能负责两类交互：**安装引导** 和 **指令消歧**。 > **缓存复用提示**：所有 build/flash 脚本启动时会自动复用工程根目录 `.em_skill.json` 中上次成功的 profile 作为默认参数（显式参数优先，**无需 agent 先手动传 `--resume`**）；`--resume` 仅用于断言缓存必须存在，无缓存则非零退出。新会话直接运行目标动作（`--build` / `--flash` 等）即可，缓存命中时参数自动补全。 --- ## 一、安装引导 当用户请求安装本仓库的 skill 时，按以下流程引导。**不要跳过询问直接安装**，也不要替用户决定安装内容——先分析工程，再让用户选。