SKILL.md
When you need to execute code safely:
- Check if RLM Runtime is installed:
- Run: which rlm or rlm --version - If not found, suggest: pip install rlm-runtime[all]
- For safe execution, use Docker environment:
- rlm run --env docker "Your task here" - Docker provides isolation from host system
- For quick local execution (trusted code only):
- rlm run "Your task here"
- Common patterns:
- Run tests: rlm run --env docker "Run pytest for the authentication module" - Validate code: rlm run --env docker "Check if the API endpoint returns valid JSON" - Complex computation: rlm run --env docker "Parse logs and extract top 5 error messages"
- View execution logs:
- rlm logs - See recent executions - rlm visualize - Launch interactive trajectory viewer
When to use:
- User asks to "run tests", "execute code", "validate implementation"
- You need to verify code behavior before suggesting changes
- Complex data processing or computation needed
- Running potentially unsafe user code (always use Docker)
When NOT to use:
- Simple code review or explanation
- Reading or editing files (use Read/Edit tools instead)
- git operations (use Bash tool instead)