What problem does it solve? AI coding sessions can execute destructive commands like rm -rf, git push --force, or DROP TABLE, and edits can spill into unrelated files. This Skill adds opt-in guardrails that warn on dangerous shell commands and confine file edits to a chosen directory. ## Core Features & Use Cases - Cautious Mode: Scans every Bash command against patterns for recursive deletion, SQL drops, force pushes, hard resets, piped remote execution, and fork bombs, emitting a stderr warning before execution. - Lockdown Mode: Restricts Edit and Write operations to a single directory tree, blocking changes outside the allowed path with an explanation. - Session-Scoped State: Stores mode and lockdown path in a per-session temp file, cleared with /safe-mode clear. - Use Case: While refactoring one API module, run /safe-mode lockdown src/api/ so the AI cannot accidentally modify tests or unrelated code, and combine it with cautious mode to catch destructive git commands. ## Quick Start Ask the AI to activate safe mode in cautious mode, or lock down edits to a specific directory such as src/api/.
