Destructive data changes
DROP, TRUNCATE and DELETE statements.
PRE-FLIGHT CHECK / 001
Pause before the
irreversible command.
Paste a command to check listed destructive patterns. Install the Claude Code hook to make this check run before Bash tool calls.
Command checks run locally
Commands stay in this browser. Submitting feedback sends your chosen answers, not the command, to Netlify Forms. No API key, terminal access or database connection.
doctor.js against your settings · 3 Confirm PreToolUse in Claude Code /hooks. Test only in a disposable project.01 / INPUT
What is about to run?
SQL, shell, Terraform and cloud CLI commands supported19 / 2,000
Try a sample
02 / DECISION
Preflight result
Waiting for a command
The result will include the risk level, matched intent and the next safe action.
BLOCK
Destructive operation detected
MATCHED SIGNALS
NEXT SAFE ACTION
03 / POLICY
What AgentGate checks
Visible pattern checks.
Review every command yourself.
Approval bypass
Automatic approval flags alongside risky commands.
Resource deletion
Terraform destroy, selected cloud deletes and broad shell deletion.
CLEAR means no listed pattern matched. It does not verify a command, inspect your environment or guarantee safety.
04 / INTEGRATE
Install the hook. Verify it loaded.
Download the CLI, Claude Code hook and setup checker. Run node doctor.js --settings /path/to/.claude/settings.json, then confirm the active hook in Claude Code with /hooks.
05 / PILOT
Need a second pair of eyes on your hook?
Tell me where the setup is failing. I’m testing a fixed-scope review for small teams: review one repository’s Claude Code hook setup, reproduce one harmless test command, and return a short findings report. No production access needed.
Pilot price hypothesis: $99 per review. The scope and price are confirmed before any work starts. This form is an enquiry, not a payment.