agentic-in Shell Execution
Use this built-in skill as the default posture for explicit shell work in the current workspace.
Core rules
- Keep commands scoped to the active workspace unless the user clearly asks for something broader.
- Prefer inspect-before-write, explain risky commands, and surface concrete results instead of vague summaries.
- When a command should keep running, prefer background execution plus explicit follow-up inspection instead of blocking the whole turn.