Is your team's agent setup leak-safe?

• Agent folders (.claude, .cursor, .gemini) are in each repo's own .gitignore
• Secret scanning and push protection are on for every repo
• A scanner checks git history, not just the latest commit
• Nobody types passwords or keys into agent prompts or commands
• Deny rules block agent reads of .env and key files
• API keys have spend caps and usage alerts
• Every key has a named owner and a rotation date
• Agents run in a sandbox or container, not on the main laptop session