>> skills/stm32-pipeline

stars: 0
forks: 0
watches: 0
last updated: 2026-03-21 16:24:14

STM32 Development Pipeline

Run the complete 4-agent pipeline for STM32 firmware development.

Pipeline: $ARGUMENTS

Execute the following agents in sequence, passing context forward:

Step 1: Environment Analysis

Launch the stm32-env-analyzer agent to understand the user's development environment.

Provide this context to the agent:

  • User's request: $ARGUMENTS
  • Scan the project for existing code, configuration files, and toolchain artifacts
  • Ask the user clarifying questions about their MCU, toolchain, and approach (HAL/LL/bare-metal)
  • Output a structured environment profile

Wait for the environment profile before proceeding.

Step 2: Documentation Research

Launch the stm32-doc-researcher agent with the environment profile from Step 1.

Provide this context to the agent:

  • The environment profile from Step 1
  • The user's original request: $ARGUMENTS
  • Search docs/ folder for relevant PDFs (reference manual, datasheet, application notes)
  • Extract register maps, configuration sequences, memory layouts, and any information needed to fulfill the request
  • Output structured documentation extractions with page references

Wait for the documentation extractions before proceeding.

Step 3: Code Generation

Launch the stm32-coder agent with context from Steps 1 and 2.

Provide this context to the agent:

  • The environment profile from Step 1
  • The documentation extractions from Step 2
  • The user's original request: $ARGUMENTS
  • Generate production-quality code following CMSIS conventions and STM32 best practices
  • Output complete, compilable source files

Wait for the generated code before proceeding.

Step 4: Code Review and Fix

Launch the stm32-code-reviewer agent with the code from Step 3.

Provide this context to the agent:

  • The generated code from Step 3
  • The documentation extractions from Step 2 (for verification against RM)
  • Review using the STM32-specific checklist (clock config, init order, interrupts, DMA, volatile, etc.)
  • Report issues by severity (Critical/High/Medium)
  • Regenerate the code with ALL fixes applied

Step 5: Present Results

After all agents complete, present to the user:

  1. Environment summary
  2. Key documentation findings
  3. Final reviewed and fixed code
  4. Any remaining recommendations or manual verification steps
    Good AI Tools