What problem does it solve? Autonomous agents and production work carry the risk of accidental destructive operations like force pushes, recursive deletions, or unintended file edits outside a target area. This Skill intercepts those dangerous actions before they execute, preventing data loss and unintended system changes. ## Core Features & Use Cases - Careful Mode: Detects destructive command patterns (rm -rf, git push --force, DROP TABLE, kubectl delete, and more) and asks for confirmation with a safer alternative before execution. - Freeze Mode: Locks all Write/Edit operations to a specific directory tree, blocking changes outside the allowed scope. - Guard Mode: Combines both protections for maximum safety when agents run in full-auto mode, with all blocked actions logged to a safety log file. - Use Case: When running an agent autonomously on a production codebase, enable Guard Mode scoped to src/api/ so the agent can read anything but only modify that directory, while destructive shell commands are blocked everywhere. ## Quick Start Ask the agent to enable safety guard in guard mode restricted to the src/api directory before starting the autonomous refactoring session.
