Feature: Safety Features & Dangerous Command Detection

[mgd43b]

Overview

Detect and warn about potentially dangerous commands before execution.

Problem

Users sometimes accidentally run dangerous commands (rm -rf, production deletions, etc.). No safety net exists.

Proposed Solution

Implement safety guardrails:

• Detect dangerous commands before execution
• "This will delete production data - are you sure?"
• Learn from your mistakes (flag similar commands you Ctrl+C'd)
• Context awareness: "You usually run this in dev, not prod"
• Configurable safety rules

Examples

$ rm -rf /
⚠️  WARNING: This command will recursively delete from root!
Are you sure? (type 'yes' to confirm):

$ kubectl delete deployment app --context=prod
⚠️  You're about to delete from PRODUCTION
   You usually test this in dev-context first
Proceed? (y/N):

Technical Considerations

• Pattern matching for dangerous operations
• Context detection (pwd, git branch, kube context)
• Machine learning from user behavior
• Performance: must not slow down normal commands

Priority

High - Safety-critical feature