메뉴
Use when building a closed STM32-to-Matlab workflow for J-Link HSS data capture, system identification, PID/LQR/MPC tuning, Kalman parameter estimation, or Matlab/Simulink code-generation feedback loops.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | bfd-matlab-mcd |
| description | Use when building a closed STM32-to-Matlab workflow for J-Link HSS data capture, system identification, PID/LQR/MPC tuning, Kalman parameter estimation, or Matlab/Simulink code-generation feedback loops. |
BFD-Kit/.learnings/ERRORS.md and/or BFD-Kit/.learnings/LEARNINGS.md.BFD-Kit/.learnings/FEATURE_REQUESTS.md.BFD-Kit/.learnings/CHANGELOG.md.Use this skill for the full loop:
python3 ./.codex/skills/bfd-project-init/scripts/bootstrap.py --project-root . --mode check
/home/xuan/matlab/bin/matlab -batch "disp(version)"
If Matlab is installed elsewhere, pass --matlab-bin <path> or set MATLAB_BIN.
python3 BFD-Kit/scripts/bfd_experiment.py capture-hss \
--elf "${STM32_ELF}" \
--symbol motor_cmd \
--symbol motor_speed \
--duration 2.0 \
--period-us 1000 \
--output-dir logs/experiments \
--experiment-name motor_step_001 \
--stimulus "step command"
The command writes capture.csv, capture.csv.meta.json, manifest.json, summary.json, summary.md, and Matlab templates under matlab/. Terminal output is compact by default.
python3 BFD-Kit/scripts/bfd_experiment.py summarize \
--dataset-dir logs/experiments/motor_step_001
Use summaries for AI review. Do not paste full CSV files unless the task explicitly requires raw samples.
python3 BFD-Kit/scripts/bfd_experiment.py matlab-run \
--dataset-dir logs/experiments/motor_step_001 \
--analysis system-id \
--matlab-bin /home/xuan/matlab/bin/matlab
matlab-run defaults to --matlab-backend auto. In auto mode, BFD-Kit first tries the official MathWorks matlab-mcp-core-server; if the MCP server is unavailable or fails before a Matlab script runs, it records the MCP error under matlab_out/matlab_mcp_error.log and falls back to matlab -batch.
Use explicit MCP mode when validating an agentic Matlab/Simulink session:
python3 BFD-Kit/scripts/bfd_experiment.py matlab-run \
--dataset-dir logs/experiments/motor_step_001 \
--analysis mcd-check \
--matlab-backend mcp \
--mcp-server /path/to/matlab-mcp-core-server \
--mcp-matlab-root /home/xuan/matlab \
--satk-root /path/to/simulink-agentic-toolkit
The MCP path uses detect_matlab_toolboxes, check_matlab_code, and run_matlab_file when exposed by the server. If --satk-root is provided, it also runs satk_initialize through evaluate_matlab_code before the dataset analysis file, enabling Simulink Agentic Toolkit model-query and MBD workflows in the same session.
Supported --analysis values:
system-id: build iddata from classified input/output columns and estimate a conservative first-order model.control: produce baseline PID, LQR, and MPC candidates from an identified model.kalman: estimate measurement R and process Q covariance candidates from selected columns.mcd-check: record Matlab/Simulink codegen toolbox availability and MCU integration rules.manifest.json lacks classification, update classification before trusting system identification output.matlab -batch only as an explicit or automatic fallback.USER/Modules or USER/APP, not CubeMX-managed directories.BFD-Kit/docs/plans/stm32-matlab-mcd-workflow.mdBFD-Kit/scripts/bfd_experiment.pyBFD-Kit/resources/matlab/templates/bfd-data-acquisitionbfd-debug-orchestratorUse when acquiring STM32 runtime data from global symbols, memory regions, stack-published local slots, or RTT channels and converting captures into structured artifacts for analysis.
Use when debugging STM32 applications with J-Link or ST-Link, including breakpoints, watchpoints, register or memory inspection, HardFault analysis, and automated fault-context capture.
Use when flashing STM32 firmware, verifying flash results, diagnosing download failures, or running repeatable CLI-based flash workflows.
Use when building a closed STM32-to-Matlab workflow for J-Link HSS data capture, system identification, PID/LQR/MPC tuning, Kalman parameter estimation, or Matlab/Simulink code-generation feedback loops.
Use when preparing, auditing, repairing, or launching SEGGER Ozone STM32 debug projects (.jdebug / .jdebug.user). Trigger this skill when Ozone shows stale Windows paths, wrong ELF/SVD/device/interface settings, missing FreeRTOS awareness, duplicated case-variant project files, stale fixed J-Link serial bindings, or when a user wants a ready-to-debug Ozone session opened from the current STM32 project.
Use when capturing STM32 RTT output through the current J-Link mainline, validating runtime behavior after flash/reset, or collecting short runtime evidence from J-Link RTT channels.