Rtos workflow
RTOS kernel engineer daily workflow router. Use when user mentions: requirements analysis, design review, code implementation, driver development, build firmware, debug crash, memory leak, performance profiling, patch review, code review, commit, push, cherry-pick, rebase, 需求分析, 方案设计, 写驱动, 编译, 调试, 崩溃分析, 内存泄漏, 性能分析, 代码审查, 提交代码, or any embedded/RTOS development task.From its SKILL.md
npx -y skills add Kaben123/rtos-engineer-toolkit --skill rtos-workflowAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.2 KB, as published. Nobody here has run it
RTOS Daily Workflow Router
Route the user's task to the correct approach based on their intent.
Prerequisites
Read .claude/rtos-knowledge/profile.md to identify the current RTOS type and project context. If this file does not exist, invoke the project-init skill first.
Identify Intent
Classify the user's request into one of five workflows:
| Workflow | Trigger signals |
|---|---|
| Requirements & Design | issue tracker, design doc, requirement, architecture, 需求, 方案 |
| Code Implementation | write driver, new feature, fix bug, build, compile, config, 写代码, 编译 |
| Debug & Analysis | crash, coredump, memory leak, performance, trace, hang, 崩溃, 内存, 性能 |
| Patch Review | review, audit, diff, check style, 审查, 评审 |
| Commit & Collaborate | commit, push, cherry-pick, rebase, 提交, 推送 |
Workflow: Requirements & Design
- Read the requirement source (issue tracker / doc / user description)
- Identify affected code modules in the current repo
- Output:
- Task breakdown (each item < 1 day)
- Risk & dependency list
- Acceptance checklist
- Ask user to confirm before proceeding
Workflow: Code Implementation
- Read
.claude/rtos-knowledge/build.mdfor build commands - Read
.claude/rtos-knowledge/coding-style.mdfor style rules - Clarify target: which subsystem, which board/target
- Follow existing coding patterns in the repo
- Mandatory checks before declaring done:
- Compiles without error (using build command from knowledge)
- Passes style checker (from knowledge)
- Matches existing code patterns
- Never generate APIs that don't exist — always grep to verify
Workflow: Debug & Analysis
- Read
.claude/rtos-knowledge/toolchain.mdfor debug tools - Follow the 3-hypothesis method:
- Parse the crash/log/trace data
- Locate source code lines using project toolchain
- Propose exactly 3 most likely root cause hypotheses
- For each, state how to verify (specific command or check)
- Ask user which hypothesis to investigate first
Workflow: Patch Review
- Read
.claude/rtos-knowledge/coding-style.mdfor style rules - Review the diff against these dimensions (report by file + line + severity):
| Severity | Meaning |
|---|---|
| Block | Must fix before merge — correctness, security, data loss |
| Major | Should fix — resource leak, missing error path, race condition |
| Minor | Improvement — naming, redundancy, readability |
| Nit | Style only — formatting, comment wording |
Kernel-specific checks:
- ISR context: no blocking calls in interrupt handlers
- Synchronization: correct primitive for the scenario
- Resource lifecycle: every alloc path has a corresponding free path
- Error paths: all fallible ops checked, resources freed on error
Workflow: Commit & Collaborate
- Read
.claude/rtos-knowledge/conventions.mdfor commit format and push model - Stage relevant files (never stage secrets or binaries)
- Generate commit message following the project's convention
- Run
pre-submit-reviewas a quality gate before push - If review passes: push using the project's push model (from knowledge)
- If review finds blocking issues: fix first, then re-review
- Never force-push; never push directly without user confirmation
Note: Use task-tracker to record the patch link after successful push.
Quality Guardrails (Always Apply)
- Verify before trust: Claude-generated APIs may not exist — grep to confirm
- No secrets in context: reject if user pastes keys/tokens
- Every change must be verifiable: answer "how do I verify this is correct?"
- Human decides architecture: propose, don't decide
- Build → Static check → Test → Device verification: this pipeline is non-negotiable