Damage Control

You can block dangerous bash commands in Claude Code by implementing PreToolUse hooks that intercept harmful patterns like rm -rf or git reset --hard before execution. Damage Control loads configurable patterns.yaml files at runtime to enforce these command restrictions automatically.

Yes, you can protect sensitive files in Claude Code by configuring path-based permissions. Damage Control uses zeroAccessPaths, readOnlyPaths, and noDeletePaths configurations to guard sensitive data and system areas from unintended modification or deletion.

To require user confirmation for risky operations in Claude Code, define ask patterns in your configuration. Damage Control triggers explicit confirmation prompts when Claude Code attempts operations matching these patterns, preventing inadvertent data loss or security breaches.

You need either Python UV or Bun to run security hooks in Claude Code. Damage Control ships with reusable hook implementations in both languages that load patterns.yaml at runtime, enforcing PreToolUse and permission hooks during command execution.

To set up security guardrails for shared Claude Code environments, install Damage Control hooks globally or at the project level, then customize patterns.yaml to cover your team's critical paths and commands. This defense-in-depth system prevents inadvertent data loss across shared setups.

A PreToolUse hook is an interception mechanism that runs before Claude Code executes commands. You need it for Claude Code security when enforcing command pattern blocking, path-based protection, and confirmation prompts to prevent dangerous operations from running.